Application of blockchain technology in teaching evaluation in applied technology universities

1. Introduction

Teaching evaluation is the main means for colleges and universities to monitor the quality of teaching and to understand the quality of teaching. Colleges and universities have been carrying out teaching quality evaluation in order to understand the teaching effect in time. Therefore, constructing an evaluation index system that meets the needs of university teaching and talent cultivation is not only the requirement of objectively and reasonably evaluating teachers’ classroom teaching and improving the quality of classroom teaching, but also a choice based on university teaching management and talent cultivation orientation [15,10]. On the basis of the traditional university classroom teaching index system, exploring the construction of the evaluation index system in line with the talent cultivation of universities of applied technology has a positive significance for the transformation of existing traditional universities and the cultivation of talents in new universities of applied technology [9].

With the rapid development of information technology, the traditional teaching evaluation method has been difficult to meet the needs of modern teaching management. Traditional teaching evaluation often has problems such as strong subjectivity, easy to tamper with data, and non-transparent process, which leads to low recognition of teaching evaluation results and cannot provide strong support for teaching decision-making [11]. Relevant government documents clearly point out that colleges and universities should comprehensively use and integrate information technology to promote the scientific, professional and objective teaching evaluation system. The implementation of this program aims to improve the accuracy and efficiency of teaching evaluation through technical means, to ensure that the teaching evaluation process is more fair and reasonable, so as to better serve the development of the teaching career and the needs of talent cultivation [7]. Therefore, exploring a new, technology-based teaching evaluation paradigm has become an urgent need in the current teaching field. Blockchain technology provides a brand new solution for teaching evaluation in universities of applied technology by virtue of its decentralization, non-tampering and transparent and open characteristics. By applying blockchain technology to teaching evaluation, the transparency of the evaluation process, the guarantee of data authenticity, and the fairness of evaluation results can be realized [13,18].

The development of blockchain technology provides technical support for colleges and universities to build a scientific teaching evaluation system. Literature [17] proposes an educational evaluation management system based on blockchain technology, and compared with other traditional evaluation systems, the proposed system can solve the drawbacks existing in the current educational evaluation process, provide a blockchain information data platform, and make the whole process of educational evaluation transparent and open. Literature [5] proposes a Collaborative Learning and Student Work Evaluation (CLSW) model supported by Blockchain Technologies (BCTs), which can improve the sustainability of the educational process in general, and the results of the polling survey show that lecturers want and are willing to teach using BCTs. Literature [23] incorporates K-means clustering algorithm and blockchain technology to design a student competency evaluation system in order to realize the whole process of monitoring the evaluation of students’ vocational competencies, objectively and effectively evaluating the students’ vocational competencies, which provides the possibility of creating an ecosystem for student competency evaluation in the future. Literature [22] introduces blockchain technology in the education and teaching of higher vocational colleges and universities, builds a big data system of individual learning and information, and creates a platform of intelligent education and teaching in higher vocational colleges and universities, which is of great significance for improving the level of teaching and evaluation in vocational education. Literature [6] indicates that blockchain technology can be applied to teaching evaluation as a motivating factor for developing learning ability, improving students’ learning and creative enthusiasm in the teaching process, and realizing the sustainable development of students’ learning process. Literature [12] tries to apply blockchain technology in cultural and creative design and education, and through empirical analysis shows that blockchain technology can effectively optimize the teaching structure and teaching management evaluation of cultural and creative design education.

In addition, literature [3] proposes a reliable and transparent student assessment and evaluation system based on blockchain, and verifies the superior performance of the proposed system in terms of assessment security, data integrity, and process efficiency through case studies and simulations, which ensures that students can receive honest and fair assessments and evaluations. Literature [16] proposed a blockchain-based online course design and evaluation methodology, which explored the teachers’ perceptions and experiences of the value of blockchain in course design by conducting a semi-structured interview survey with five teachers, and the results showed that the proposed methodology could improve the match between blockchain and the course, the quality of teaching and learning, and the trust of all parties involved in online education. Literature [8] proposes an educational blockchain technology based on learning outcomes and meta-diplomas, which can convert student performance evaluation results to job competency evaluation results, and feed back the reaction of student competency evaluation into the curriculum for continuous improvement of the curriculum. Literature [4] first investigates teachers’ desire and willingness to introduce area chain technology in education, then proposes a collaborative learning and student work evaluation model based on blockchain technology, and finally verifies the feasibility and effectiveness of the proposed model through case implementation. Literature [1] briefly analyzes the key technology of blockchain, the main content of teacher evaluation work and the non-functional requirements of the system, and constructs a teacher evaluation system for colleges and universities based on blockchain technology on this basis, and verifies through performance testing that the system can effectively deal with practical problems such as high pressure on the collection of evaluation information, data wastage, complex statistical operations, and inaccurate reflection of evaluation results. Literature [2] creates a new learning model based on the student-centered iLearning blockchain framework, which can be adapted to the specific professional needs of the employment sector, and can be applied to the teaching evaluation of all training institutions for assessing and measuring the skill level of students. Literature [21] applies blockchain technology with decentralized and safe and reliable characteristics to the teaching quality evaluation and guarantee system of universities to construct a teaching evaluation system with intelligence, fairness and accuracy to promote the construction and better development of university disciplines.

This paper analyzes the coupling relationship between blockchain technology and teaching quality evaluation in colleges and universities. Blockchain technology is utilized to build a four-faceted university teaching quality evaluation technical architecture that forms the data layer, network resource layer, consensus access layer, and application layer. The improved PBFT consensus algorithm is used to maintain the data layer, the blockchain P2P architecture carries the infrastructure function, and the POS mechanism and decentralized applications (DApps) realize the unified management of teaching quality evaluation in colleges and universities. According to the teaching characteristics of universities of applied technology, we construct the structure of practical teaching evaluation system of universities of applied technology, incorporate the improved PBFT algorithm, and propose the mode of “off-chain storage + on-chain sharing” for practical teaching evaluation data of universities of applied technology. The operational feasibility of the improved consensus algorithm is analyzed and brought into the practical teaching evaluation system of the University of Applied Sciences for overall system performance testing. Construct the teaching quality evaluation index system of universities of applied technology based on blockchain technology, and verify the suitability of blockchain technology and higher education evaluation.

2. Feasibility of blockchain technology for teaching evaluation in universities of applied sciences

2.1. The coupling relationship between blockchain technology and teaching quality evaluation in colleges and universities

Blockchain technology is a kind of in order to solve the centralized relational database storage of low efficiency, high cost, low data security and other problems, and the use of decentralized distributed database to store information data, with the help of data encryption, proof of workload, timestamping technology and other means of recording and disseminating the information, in the no trust premise of the storage nodes distributed network of decentralized credit peer-to-peer transmission, collaboration and coordination technology [14,20,19].

At present, blockchain has been applied in many fields, such as taxation, logistics, medical and other fields. It is not difficult to find that the technical characteristics of blockchain are strongly coupled with the high performance teaching quality evaluation system.

The coupling relationship between blockchain technology and university teaching quality evaluation management is shown in Table 1.

As can be seen from the table, blockchain technology and university teaching quality rating management have a highly coupled relationship. It is mainly manifested in the aspects of, decentralized technology, trust-based infrastructure, smart contract, coordinated sharing mechanism, asymmetric encryption algorithm and time tracing. The teaching quality rating system of colleges and universities created by using blockchain technology can improve the efficiency and accuracy of teaching quality rating management work in colleges and universities.

2.2. Technical architecture of teaching quality evaluation in universities based on blockchain technology

In view of the highly coupled relationship between blockchain technology and college classroom teaching quality evaluation system, the system model framework of college teaching quality evaluation based on blockchain technology can be constructed. The framework of teaching quality evaluation system of colleges and universities based on blockchain technology is established by utilizing blockchain distributed data storage, consensus mechanism, etc. including data layer, network layer, shared access layer, application layer, etc., and the framework of teaching quality evaluation of colleges and universities based on blockchain technology is shown in Figure 1.

2.2.1. Data layer

As the bottom layer of the technical architecture in the university teaching quality evaluation system, the data layer is also the most central layer, which is the core of the whole blockchain work.

The data in it comes from expert supervisors, peer evaluations and students and other subjects receiving education. These data include the type of data, the subject of classroom teaching quality evaluation of the subject’s subject category, teaching time and so on, but also includes the evaluation time, evaluation location and other basic information. These data are real and can be shared, and the historical data fingerprint storage based on blockchain ensures that the data are valid and real. In traditional databases, the background can artificially create, read, edit, modify and delete operations on the data. Blockchain design is more simplified, removing the modification and deletion operations on data. Managers in the evaluation of teaching quality in higher education can only add data to the block, and all confirmed data entering the block will not be able to be changed.

2.2.2. Network resource layer

The infrastructure will rely on the blockchain P2P architecture, no central server, each node even if the client is also the server, even if the attack, any one server collapses, will not affect the overall service.

P2P network has strong scalability, even with thousands of nodes, but still very efficient. It is also highly fault-tolerant, so that even if a node leaves, joins, or leaves during operation, the reliability of the system will not be affected. Distributed ledger means that the data storage of blockchain is distributed storage, and the storage ledger of each node is completely consistent, which is guaranteed by the blockchain P2P communication mechanism.

2.2.3. Consensus access layer

The POS mechanism is based on the node with the highest interest rather than the highest arithmetic power in the system to obtain the block bookkeeping rights.The POS consensus process relies only on the internal coin age and interest and does not need to consume external arithmetic power and resources, which fundamentally solves the problem of wastage of arithmetic power in the POW consensus, and can shorten the time to reach a consensus to a certain extent. The verification pool POA algorithm can improve the network topology compared to other algorithms, maintain the proportion of online nodes, require less transaction fees and reduce the energy loss in the consensus algorithm process.

2.2.4. Application layer

The blockchain application layer enriches the entire blockchain ecosystem by developing decentralized applications (DApps), i.e., by calling the interfaces of the protocol layer and the smart contract layer in order to adapt to all kinds of application scenarios of the blockchain, and provide various services and applications for users.

Among the management of teaching quality evaluation in colleges and universities, the teaching affairs management system of colleges and universities carries out unified management of teachers’ teaching quality and scientific research performance. Based on the blockchain distributed ledger, the data is divided into various blocks for management and query, and at the same time, based on the smart contract data tampering technology, the final report of the evaluation of the quality of college classroom teaching is all encapsulated in a block, which cannot be manipulated by others. Using time stamp technology, the final data and reports can be traced back, and the reward mechanism for teachers is absolutely “traceable”.

3. Evaluation of practical teaching in applied technology universities based on blockchain technology

3.1. Construction of evaluation index system

Evaluation index system is the basis for ensuring the quality of practical teaching, and it is a comprehensive description of the activities and related factors of practical teaching. The evaluation index system evaluates the quality of practical teaching and the teaching process, and optimizes the practical teaching activities and related factors according to the evaluation results, so as to build a set of effective evaluation index system. In the construction of the index system, we should take into account the scientific, integrity and complementarity of each link, and any weak link may lead to loopholes in the whole index system, making the links disconnected. Therefore, it is necessary to ensure that the personnel involved in each link can supervise and promote each other.

According to the construction principles of the actual teaching evaluation index system, from the starting point of promoting the quality of practical teaching and improving students’ practical innovation ability, the practical teaching quality evaluation index system was formulated on the basis of fully summarizing experience. The block diagram of the structure of the practical teaching on indicator system of the University of Applied Sciences is shown in Figure 2, which is based on the talent cultivation objectives, carries out the practical teaching curriculum system, and provides teaching guarantee for the content of practical teaching and the process of practical teaching.

3.2. Blockchain technology improvements

In the process of teaching evaluation in universities of applied technology, the data for the evaluation of the experimental process are not secured, and there are problems such as the credibility of the data. That is, there are problems such as students’ denial, tampering and falsification of data related to the experimental process. The experimental process evaluation data is packaged and uploaded to the blockchain, and a consensus needs to be reached between the nodes when uploading to the blockchain. In order to improve the efficiency of consensus, Practical Byzantine Fault Tolerance (CBFT) algorithm based on trust degree is proposed.PBFT algorithm suffers from the problems of high consensus latency and low throughput. To address these problems, the master node selection strategy and consistency protocol in the consensus process are optimized with focus on trust degree to improve the efficiency of reaching consensus.

3.2.1. PBFT consensus algorithm

1. PBFT consensus algorithm flow

   PBFT algorithm initialization all nodes need to synchronize views to ensure that nodes are in the same working state. The initial view number 0 starts. After the completion of all nodes view same, the master node \(p\) is selected, and the selection formula is as follows: \[\label{GrindEQ__1_} p=v\, mod\, n , \tag{1}\] where \(v\) is the view number. \(n\) is the node number.

2. PBFT algorithm mechanism

   The view change mechanism is activated when a node times out due to network and other reasons to ensure the activity of the system. There is a timer in each node, which is started after receiving a request, and the node is considered to be timeout if no response is received to the relevant request within a period of time, preventing the node from waiting indefinitely for the request to be executed.

   When the timeout protocol is triggered, the backup node sends a view change message to all nodes. When the backup node receives \(2f+1\) view switching requests, it initiates a view switch. The view number plus 1 becomes the new view.

   The master node calculates it by the following formula: \[\label{GrindEQ__2_} p=(v+1)modn , \tag{2}\] where \(v\) is the view number before the timeout. \(n\) is the number of the node.

During the view change process, the node accepts only CHECKPOINT, VIEW-CHANGE and NEW-VIEW messages and other types of messages are in temporary waiting state.

All the nodes in the PBFT algorithm need to record the successful get consensus transaction messages in their own logs, and as the volume of transactions increases, the node storage content will become larger and larger. The stored messages may become invalid after some time. In order to clear the meaningless messages and save the memory of the nodes, garbage collection mechanism is used in PBFT algorithm.

The state of the node at the time of request execution during each node proving process is called as checkpoint. A verified checkpoint is called a stable checkpoint. The process of checkpoint proving is as follows: the node generates the checkpoint and sends a message to all the slave nodes. The format of the message is \(<CHECKPOINT,n,d,i>\), where \(n\) is the serial number of the last successful transaction of the node and \(d\) is the summary of the state. After receiving the checkpoint message, other nodes record the node post message in the log. A checkpoint stop is proved by the fact that the node has received \(2f+1\) messages with the same status information from different nodes. In order for a node not to stop working while waiting for the checkpoint proof, it is necessary to set the associated thresholds waterline and high and low values \(h\) and \(H\). The low value of the waterline \(h\) is the same as the serial number of the last transaction, while the high value of the waterline is computed by the following formula: \[\label{GrindEQ__3_} H=h+k , \tag{3}\] where \(k\) is the large enough to be set.

3.2.2. CBFT algorithm

Aiming at the problem of high overhead of PBFT view switching and randomly selecting the master node, the master node is selected by weighted voting, which can improve the consensus efficiency. The consensus flow of CBFT, a consensus algorithm based on reputation model designed in this paper, is shown in Figure 3.

It is assumed that in the network, it contains honest nodes and malicious nodes. Honest nodes are able to respond correctly to messages. While malicious nodes behave arbitrarily, including not responding to messages or responding with timeout, or even propagating error messages. Malicious nodes suffer from network attacks or initiate attacks, each node is independent and failure of one node does not affect other nodes. In this protocol, reliable nodes are selected to join the consensus group through reputation model. The consensus is completed by the consensus group in the execution cycle rather than all nodes participate in the consensus, the main node of the committee members is responsible for leading the consensus is called the leader node, and the other nodes in the committee are called the consensus nodes. The other nodes of the system other than the committee nodes are responsible for synchronizing the network data and are called candidate nodes. Candidate nodes have different probability of joining the consensus group based on their reputation value and change the consensus group based on the performance of the nodes after the execution cycle. If malicious nodes gather in the consensus group it will lead to the destruction of the network and no consensus can be successful. After the specified number of consensus rounds, node trustworthiness is judged by node reputation characteristics and other characteristics. If a node is judged to be untrustworthy, it will be downgraded and will not be able to enter the consensus group in the following rounds.

1. Consensus node reward and punishment mechanism

   In this paper, the reputation model corresponds to the reward function and punishment function as follows: \[\label{GrindEQ__4_} R_{k+1} (n)=R_{k} (n)+r_{sk} \cdot r_{k+1} (n) , \tag{4}\] \[\label{GrindEQ__5_} R_{k+1} (n)=R_{k} (n)-p_{sk} \cdot r_{k+1} (n) , \tag{5}\] where \(n\) denotes the node number, \(R_{k+1} (n)\) denotes the reputation value of the node after \(k+1\) rounds of consensus, \(r_{sk}\) is the reward factor, \(p_{sk}\) is the penalty factor, and \(r_{k+1} (n)\) denotes the reputation value calculated based on the history of the node.

   In this paper, node participation, consensus success rate, and number of blocks generated are selected as the judgment criteria, and these indicators affect the calculation of node reputation value. Different weights are set for different indicators, and the obtained reputation value can objectively evaluate the node’s consensus performance and provide different degrees of rewards and punishments for different nodes.

   It refers to the frequency of node participation in consensus in the system, reflecting the degree of node initiative. Node participation is expressed by the following formula: \[\label{GrindEQ__6_} f_{1} =\frac{r_{n} -r_{avg} }{r_{total} } , \tag{6}\] where, \(r_{total}\) denotes the total number of consensus rounds, \(r_{n}\) denotes the number of consensus rounds of the node numbered \(n\), the average of the number of times all the nodes in the system participate in consensus is denoted by \(r_{avg}\), and \(m\) is the total number of nodes in the system: \[\label{GrindEQ__7_} r_{avg} =\frac{\sum _{i=0}^{m}r_{i} }{m} . \tag{7}\]

   Consensus success rate indicates the proportion of the number of successful nodes participating in the consensus to the total number of consensus, which is expressed by the following consensus, where \(r_{ns}\) indicates the number of successful node consensus number \(n\): \[\label{GrindEQ__8_} f_{2} =\frac{r_{ns} }{r_{n} } . \tag{8}\]

   In the consensus process, if the proportion of malicious nodes in the consensus group members does not exceed the system threshold, then it is able to generate blocks normally, in the PBFT consensus algorithm the system threshold is 1/3. The number of blocks generated by a node is then defined as the number of times a node has succeeded in the consensus in the committee, and the following formula is used to define the number of blocks generated per unit of time, and \(T\) is the consensus time: \[\label{GrindEQ__9_} f_{3} =\frac{r_{ns} }{T} . \tag{9}\]

   The corresponding weights of the three indicators are \(\omega _{1} ,\omega _{2} ,\omega _{3}\). The size of the weights is related to the importance of the indicators, and the reward and penalty functions become the following equations: \[\label{GrindEQ__10_} R_{k+1} (n)=R_{k} (n)+r_{sk} \cdot \sum _{i=1}^{3}w_{i} f_{i} (n) , \tag{10}\] \[\label{GrindEQ__11_} R_{k+1} (n)=R_{k} (n)-p_{sk} \cdot \sum _{i=1}^{3}w_{i} \left(1-f_{i} (n)\right) . \tag{11}\]

   Through the above reputation value calculation formula, the node’s reputation value change magnitude in this round of consensus is determined based on the node’s participation, consensus success rate, and the number of blocks generated per unit time, and the node’s score is positively correlated with its performance.

2. Reputation model based consensus node selection

   Nodes get public-private key pairs through Gen, each node owns the public key of other nodes, and the reputation value is obtained through the reputation model, which prevents attacks on the consensus group due to the unpredictability of the VRF results. Suppose the system has \(n\) node, node public key VPK, VSK, and reputation value \(R\) is represented as follows: \[\label{GrindEQ__12_} VPK=\{ vpk_{1} ,vpk_{2} ,\cdots ,vpk_{n} \} , \tag{12}\] \[\label{GrindEQ__13_} VSK=\{ vsk_{1} ,vsk_{2} ,\cdots ,vsk_{n} \} , \tag{13}\] \[\label{GrindEQ__14_} R=\{ r_{1} ,r_{2} ,\cdots ,r_{n} \} . \tag{14}\]

   The probability of a node being able to join the consensus committee is related to its own reputation value, the higher the reputation value, the higher the probability of being selected. In this paper, nodes are divided into different sub-nodes, the number of sub-nodes is determined by the reputation value, and the probability of each sub-node being selected is expressed by the following formula. \(c\) is the number of committee nodes: \[\label{GrindEQ__15_} p=\frac{c}{\sum _{i=1}^{n}r_{i} } . \tag{15}\]

The interval [0, 1) is divided into consecutive subintervals \(l^{j}\), if the reputation value of \(r_{i}\) node embraces the number of child nodes is \(r\), the probability of selecting \(k\) child nodes from the node’s children obeys a binomial distribution \(B(k;r,p),hash_{i} /2^{hashlen}\) in the interval \(I^{j}\) means that the node has \(j\) child nodes selected, and \(hash_{i}\) means that the node’s hash value obtained through a verifiable random function, the number of times each node broadcasts the hash value, has been selected \(j\), and the proof, the other nodes The verification is passed to determine the final consensus group members: \[\label{GrindEQ__16_} I^{j} =\left[\sum _{k=0}^{j}B (k;r,p),\sum _{k=0}^{j+1}B (k;r,p)\right) , \tag{16}\] \[\label{GrindEQ__17_} B(k;r,p)=\left(\begin{array}{c} {r} \\ {k} \end{array}\right)p^{k} (1-p)^{r-k} , \tag{17}\] \[\label{GrindEQ__18_} \sum _{k=0}^{r}B (k;r,p)=1 . \tag{18}\]

The seedseed is public for each round of consensus nodes, which is obtained and broadcasted by the consensus committee in the previous round through Eq. (18), and the random generation of the seed can avoid the seed being controlled by the adversary: \[\label{GrindEQ__19_} seed_{r} =H(seed_{r-1} ||r) . \tag{19}\]

In this paper, the malicious node behavior changes dynamically, including timeout response, non-response, and sending forged messages.

In this paper, the view switching process in PBFT consensus protocol is optimized. A new master node is elected by voting, and the node’s vote is calculated by the following formula: \[\label{GrindEQ__20_} V_{total} (j)=\alpha \cdot R(j)+\beta \cdot \left(\frac{1}{N} \sum _{i=1}^{N}W_{i} V_{ij} \right) , \tag{20}\] where, \(V_{total} (j)\) is the total number of votes obtained by node \(j\). \(\alpha\), \(\beta\) is the regulation parameter and the number of nodes in the committee is \(N\). \(W_{i}\) denotes the voting weight of node \(i\), and \(V_{ij}\) denotes the vote of node \(i\) for node \(j\) with a value of 0 or 1.

3.3. Teaching evaluation of applied technology university based on blockchain technology

3.3.1. Models of blockchain technology in higher education evaluation

The digital trust mechanism of blockchain ensures the authenticity of data in higher education evaluation. Distributed sharing, whole-process recording and other characteristics can greatly improve the efficiency of higher education evaluation and enhance the function of evaluation. Decentralization and smart contract technology can promote the diversification of evaluation subjects and evaluation standards and help the effective implementation of higher education evaluation reform.

However, the realistic application of blockchain technology in higher education evaluation is a systematic project that requires the participation of multiple subjects such as the government, society, schools, teachers and students. It is constructed hierarchically according to node power, builds data catalogs and encapsulates smart contracts according to evaluation standards, etc. Starting from reality, this paper proposes the application model of blockchain technology in higher education evaluation.

Referring to the construction method of governmental blockchain, the data distributed sharing technology of blockchain is utilized to establish the mode of “off-chain storage + on-chain sharing” of data.

In the “on-chain + off-chain” system, the evaluation object updates and maintains the data in real time under the original data directory according to the daily educational activities, which reduces the workload of centralized data collation during the evaluation period. During the evaluation process, the data processing process on the blockchain is opened, and the application, authorization, confirmation, sharing and use of data by both parties are automatically executed, simplifying the work procedure. Finally, the evaluation results and feedback are open and transparent, which is conducive to the use of the results by evaluation subjects and promotes the development of education.

In the “on-chain + off-chain” system, education data are stored locally, and data certificates and data catalogs are uploaded to the chain. Blockchain technology is not only used to verify, supervise and manage the process of data sharing, but moreover utilizes the attributes of blockchain’s full-process traceability and non-tampering. It requires very low data storage capacity and is relatively easy to apply and promote. In addition to school evaluation, the system can also be used for student evaluation and teacher evaluation, among others.

3.3.2. Operational feasibility

1. Algorithm Evaluation

   In order to make an evaluation of the improved PBFT and derive the possibility of the implementation of the improved PBFT algorithm, the scoring formulas for the PBFT and the improved PBFT (CBFT) consensus algorithms are designed in this paper respectively, as shown in Eqs. (21) and (22): \[\label{GrindEQ__21_} Score_{pbft} ={\rm a}^{*} tps_{i} +b^{*} dynamic_{i} +c^{*} energy_{i} , \tag{21}\] \[\label{GrindEQ__22_} Score_{cbft} ={\rm a}^{*} tps_{j} +b^{*} dynamic_{j} +c^{*} energy_{j} , \tag{22}\] where \(\; tps_{i} ,tps_{j}\) is the throughput of PBFT and CBFT respectively, \(dynamic_{i} ,dynamic_{j}\) is the dynamics, \(\; energy_{i} ,energy_{j}\) is the energy consumption, and a, b, and c are their corresponding weights.

   In the consensus algorithm scoring formula, the scoring formula for system \(tps\) is shown in (23): \[\label{GrindEQ__23_} tps=\frac{Sendrate}{(t_{sys} +t_{cons} )^{*} N^{*} (N-1)^{*} Blocksize} , \tag{23}\] where Sendrate is the request sending rate, \(t_{sys}\) and \(t_{cons}\) are the system delay and consensus delay, N is the number of nodes involved in the consensus, and Blocksize is the bandwidth consumed by a consensus. Sendrate, \(t_{sys}\) and Blocksize are kept constant, the \(t_{cons}\) of CBFT is 1/3 of PBFT and the number of nodes N is also 1/3 of PBFT, in the case of a certain number of nodes, the packet delay is almost negligible with respect to the consensus delay, so it can be concluded as shown in Eq. (24): \[\label{GrindEQ__24_} tps_{j} \approx 9^{*} tps_{i} \Rightarrow tps_{j} >tps_{i} \tag{24}\]

   The dynamic is calculated as shown in Eq. (25), where \(Kmeans_{j}\) is the initialized dynamic score of CBFT, \(Vote_{j}\) is the group voting dynamic score, and \(Stage_{j}\) is the dynamic score of \(j\) modifying the three-phase protocol into a two-phase protocol. Whereas PBFT is static with fixed number of nodes, so the dynamics score \(dynamic_{i}\) of PBFT is 0. It can be easily concluded that CBFT is more dynamic than PBFT: \[\label{GrindEQ__25_} dynamic_{j} =\sum _{j=1}^{N}\mathop{\scriptscriptstyle\leftharpoondown}\limits^{\displaystyle\rightharpoonup} Kmeans_{j} \mathop{\scriptscriptstyle\leftharpoondown}\limits^{\displaystyle\rightharpoonup} *Vote_{j} \mathop{\scriptscriptstyle\leftharpoondown}\limits^{\displaystyle\rightharpoonup} *Stage_{j} >0\Rightarrow dynamic_{j} >dynamic_{j} . \tag{25}\]

   The energy is calculated as shown in Eq. (26).The number of consensus nodes involved in CBFT is about 1/3 of PBFT, and the consensus energy consumption is roughly 1/3 of PBFT, so the energy score of CBFT is triple that of PBFT. i.e., \[\label{GrindEQ__26_} energy_{j} =3*energy_{i} \Rightarrow energy_{j} >energy_{i} . \tag{26}\]

   In summary, the following conclusions can be drawn as shown in equation (27): \[\label{GrindEQ__27_} \left\{\begin{array}{l} {tps_{j} >tps_{i} } \\ {dynamic_{j} >dynamic_{j} \Rightarrow Score_{cbft} >Score_{pbft} } \\ {energy_{j} >energy_{i} } \end{array}\right. \tag{27}\]

   Therefore the improved PBFT consensus algorithm scores higher than the traditional PBFT consensus algorithm, so the improvement strategy proposed in this paper is theoretically feasible and provides a theoretical basis for the subsequent implementation of the improved PBFT algorithm.

2. Analysis of the improved algorithm

   For the improvement scheme of consensus node proportion control based on average stability proposed in this paper for the PBFT algorithm, the implementation and related experiments were carried out. On the computer with CPU AMD R7 5800H and operating system Window 10 64bit, the simulation environment was constructed through JAVA language. The communication between the nodes is set to have a random delay of 5\(\mathrm{\sim}\)20ms to verify the performance and related characteristics of the improved algorithm. As the consensus process proceeds, the changes of the nodes in this improved scheme are recorded and analyzed with related data. And its performance is compared and analyzed in terms of single consensus delay and other aspects. Finally, based on different scenarios, the optimization effect is compared with other improvement schemes to verify the adaptive nature of this algorithm.

This algorithm mainly takes the average stability as the core control index. In the selection of experimental conditions, in order to be close to the actual use scenarios of the real alliance chain. That is, the probability of malicious behavior is different for different nodes in the network, so all the nodes are divided into 3 parts. Through the random number generator, the probability of malicious behavior of each node in each part is controlled as 90%, 60% and 20% respectively. At the same time, the total number of nodes making malicious behaviors in each round of the system consensus process is required to remain constant. In order to visualize the CBFT algorithm in the consensus process of consensus node control. Take the total number of nodes as 20 as an example, set the number of malicious nodes as 6. Under this experimental condition, using the CBFT consensus algorithm to carry out 20 rounds of consensus, the change of its flat stability and the actual number of nodes participating in the consensus in each round is shown in Figure 4.

It can be learned that at the initial moment of the first round of consensus, since the stability coefficient of all nodes is 100, the average stability is 100, and the corresponding number of consensus nodes is 20, and all nodes participate in the first consensus process.

As the number of consensus rounds increases, the average stability increases and the corresponding number of nodes in the consensus cluster decreases. When the consensus proceeds to the 20th time, the nodes in the consensus cluster remain at 21. It can be analyzed that when the number of malicious behavior nodes is not excessive, the number of node clusters participating in the consensus will eventually be controlled at about 0.6 of the total number.

At the same time, the number of malicious nodes in the system will also have a certain impact on the decline rate of the number of nodes in the consensus cluster. The total number of nodes with malicious behavior in each round is set to 0, 1, 2, 3, 4, 5, 6, respectively, and the changes in the average stability of its 20 rounds of consensus process are observed and recorded, and the changes in the average stability of different numbers of malicious nodes are shown in Figure 5. If the number of malicious nodes in each round is less, the corresponding score reduction operation is less, thus the rate of flat stability is faster, and the number of consensus nodes to reach a stable state requires fewer rounds.

The time required from the beginning to the end of a single consensus, i.e., the consensus delay, can intuitively reflect the performance of the algorithm.

Therefore, the experiment compares the consensus delay generated by the single consensus of CBFT algorithm and PBFT algorithm to compare the consensus efficiency of the two algorithms.

The number of nodes making malicious behavior per consensus round is specified as \(\left(N-1\right)/3\), and \(N\) represents the total number of nodes. According to the different values of \(N\), respectively, carry out a number of experiments. Record the consensus delay required for each round of consensus after it reaches a stable state, and take the average of multiple times as the comparison result, the comparison of algorithm consensus delay is shown in Figure 6.

Due to the increase in the total number of system nodes, the consensus latency of both PBFT algorithm and CBFT algorithm rises significantly. However, due to the fact that the CBFT algorithm employs a proportional control of consensus nodes based on flat stability, which reduces the number of consensus nodes to a certain extent, the single-round consensus latency is improved. When the number of nodes reaches 100 times, the consensus latency of CBFT algorithm and PBFT algorithm is 84.040ms and 93.563ms, respectively. and due to the abandonment of polled view switching and the adoption of the scheme based on the stability coefficient to control the election priority of the master node, so that the probability of view switching is greatly reduced, therefore, under a variety of experimental conditions where the total number of nodes of the system takes on different values, the single-round consensus latency of CBFT algorithm under various experimental conditions with different values of the total number of nodes in the system, the single consensus delay of the CBFT algorithm is reduced.

3.3.3. Teaching evaluation system testing

CBFT is utilized in the practical teaching evaluation framework of the University of Applied Sciences in order to improve the system’s transaction throughput, transaction latency, etc., in order to ensure the user’s experience and efficiency of use.

In this section, the key performance of the practical teaching evaluation system of the University of Applied Sciences is tested for performance. The students of School A are taken as the evaluation object of the system and are tested using the specialized Caliper testing tool.Caliper is an open source blockchain performance testing tool that allows users to test different Hyperledger networks with predefined test cases and obtain a set of performance test results.

The working resource consumption test of the evaluation system based on CBFT algorithm or PBFT algorithm is shown in Table 2.

The table shows the system resource consumption information of the practical teaching evaluation system of the University of Applied Sciences when the faculty user sends 1000 transaction requests. Under the high concurrent requests, this system still can use low resources to ensure that the user’s information is properly stored in the blockchain system.

In addition, the same test was performed on the Hyperledger network system using PBFT. The main data gap is with the orderer node, the work of this node is mainly to carry out consensus and sorting operations on transactions, it can be seen that the use of PBFT in the same system will consume a lot of system resources to carry out calculations in the consensus operation, which will consume too much CPU resources, and due to the fact that collecting the packets of the consensus process of the other nodes will take up a lot of memory resources, the system using PBFT will consume a large amount of memory resources, compared to using CBFT will ensure the energy efficiency of the system and consume relatively low memory and CPU resources.The average memory consumption and average CPU consumption of the CBFT algorithm in Peer0, Peer1, and Peer2 are less than that of the PBFT algorithm.The average CPU consumption of the CBFT algorithm and the CBFT algorithm are 13.4%, 18.5%, and 18.5%, respectively. .

Through the above tests, it is proved that this teaching evaluation system realizes the high transaction throughput and low system resource consumption of the blockchain-based management system by introducing the super ledger blockchain network.

In summary, the blockchain storage of teaching data of the University of Applied Sciences guarantees the transparency and tamper-proofness of the data. And because the underlying blockchain module in this system is based on the CBFT algorithm for the consensus process of the blockchain, it improves the transaction throughput of the system and saves the consumption of system resources, which ensures that the evaluation users can use this practical teaching evaluation system of the University of Applied Sciences smoothly and can trust that the data in the system has not been tampered with.

3.4. Teaching evaluation applications

According to the principle of indicator construction, the teaching quality evaluation index system of applied technology universities based on blockchain technology is derived. Among them, the five first-level indicators are teaching externality, teaching guarantee, teaching feedback, teaching empathy, and teaching practice. And 18 second-level indicators are divided according to the specific teaching content, teaching objectives, teaching syllabus and teaching methods of applied technology university majors.

In order to further ensure the accuracy of the above indicators, firstly, with reference to the principle of 1:1:1, the opinions of 18 industry experts, 18 front-line teachers and 18 students were collected, and then after repeated collation and statistics, relatively consistent results were obtained. Combining the above scoring indicators and the opinions collected from experts, teachers and students, the judgment matrix between the indicators is derived Z. The data related to the judgment matrix of the first-level indicators are normalized and then tested for consistency. Finally, the weight value of each indicator in the matrix is multiplied by the weight value of the corresponding criterion level to arrive at the total weight of all indicators in the teaching quality evaluation system of universities of applied technology. The results of the weights of the teaching quality evaluation indicators of applied technology universities based on blockchain technology are shown in Table 3.

Among the five first-level indicators in the teaching quality evaluation system of universities of applied technology based on blockchain technology, the weights of teaching externality, teaching security, teaching feedback, teaching empathy, and teaching practicability are 0.1841, 0.2057, 0.1871, 0.1824, and 0.2407, respectively.The weights of teaching practicability and teaching security account for a high percentage, which is higher compared with other indicators to enhance the have a greater impact.

Among the five first-level indicators, the weights of A1, B3, C2, C3, D1, D2 and E2 were all greater than 0.3, with the weights of “C3 teachers can establish a way for students to give feedback on teaching” and “D2 teachers can guide and understand students’ special learning needs”, which were the largest, with 0.42 and 0.41, respectively. The University of Applied Sciences focuses on practical operations, emphasizing student needs and student opinions.

Overall, in this process of constructing the teaching quality evaluation system of applied technology universities based on blockchain technology, five primary indicators and 18 secondary indicators are synthesized. Then, further indicator weights are adopted for each indicator to determine. The evaluation system is finally concluded to be a multi-factor and multi-indicator system, which effectively realizes the innovation of the traditional evaluation system with the help of blockchain technology.

4. Conclusion

This paper constructs the practical teaching evaluation index system of applied technology universities, and utilizes blockchain technology to solve the problems of data security and data credibility in the process of practical teaching evaluation. PBFT consensus algorithm is proposed and algorithm improvement is carried out to analyze the application of CBFT algorithm in teaching evaluation of universities of applied technology.

1. Design the scoring formulas of PBFT and improved PBFT consensus algorithms respectively, in the case that the sending rate, system delay, and bandwidth consumed by one consensus remain unchanged, the consensus delay of CBFT is 1/3 of that of PBFT, and the number of nodes N is also 1/3 of that of PBFT.In the algorithmic comparison and analysis, due to the increase of the total number of system nodes, the consensus of both PBFT algorithm and CBFT algorithm latency of both PBFT algorithm and CBFT algorithm rise significantly. However, the consensus delay of CBFT algorithm is significantly less than that of PBFT algorithm, and the consensus delay of a single round is improved. When the number of nodes reaches 100 times, the consensus delay of CBFT algorithm and PBFT algorithm is 84.040ms and 93.563ms respectively.

2. In the operation of the teaching evaluation system of the University of Applied Sciences, the use of the PBFT algorithm consumes a large amount of system resources for the calculations in the consensus operation, while the CBFT algorithm ensures the energy-saving of the system, which occupies relatively low memory and CPU resources. It accelerates the operation of the teaching evaluation system of universities of applied technology and improves the efficiency of teaching evaluation.

3. Construct the teaching quality evaluation index system of applied technology universities based on blockchain technology, and calculate to get the comprehensive weight of each index. In the teaching feedback dimension, the corresponding weights of “C2 teachers’ teaching content has the cutting-edge perspective of blockchain technology” and “C3 teachers can establish a way for students to give feedback on teaching” are 0.34 and 0.42 respectively, which are of greater influence. Comprehensive weighting of indicators in each dimension can further illustrate the coupling relationship between blockchain technology and the evaluation of teaching quality in universities of applied technology, and deepen the systematic evaluation of teaching in universities of applied technology.