Research on College English Teaching Quality Assessment Method Based on K-Means Clustering Algorithm

. The evaluation of college teachers’ teaching ability is very important. Currently, the indicators for evaluating the quality of college English teaching are unclear and insufficient. This paper evaluates the quality of university classroom teaching from two aspects: students’ learning effect and teachers’ teaching work. This paper employs the K-means algorithm to analyze the relationship between the indicators in the evaluation model and teachers’ teaching ability, finds out the specific factors that affect teaching activities, and guides the implementation of teachers’ teaching work. At the same time, the K-means model is used to evaluate students’ learning effect, identify the relationship between the indicators in the model and teachers’ teaching ability, and find out the specific factors that affect teachers to guide the implementation of teachers’ teaching work. Experiments show that the method proposed in this paper can solve the problem that the evaluation indicators of traditional evaluation methods are not clear and insufficient and can be better applied to teaching evaluation.


Application of Clustering Algorithm in Teaching
Evaluation. Because the English teaching capability assessment is a more restrictive factor, quantitative tests and analysis on the need for English teaching level construct a model of parameter constraint English teaching level and large data analysis model [1].
Because information processing and big data analysis technology have important applications in teaching evaluation and information resource scheduling. e superiority of semisupervised clustering mainly lies in that a small amount of supervised sample information can be used when clustering unlabeled samples [2,3]. erefore, how to make better use of the domain knowledge contained in the annotated samples for analysis has become a research hotspot [4].
e former uses labeled sample information to constrain the clustering process and nally obtains an appropriate partition result.
e speci c methods are as follows: modify the objective function of clustering to meet the given constraints, follow the constraints in the process of clustering, use the labeled sample information to initialize the clustering parameters, and restrict the division of data in the process of clustering [5]. e latter is based on a distance measure function for clustering. e distance measure function used in this method is a more reliable distance measure function obtained by learning label data [6].
Most semisupervised clustering approaches are currently built from conventional clustering algorithms, which augment the supervised sample information provided by conventional clustering algorithms. Among these, the K-means technique has been extended to the semisupervised domain as a simple and e cient clustering algorithm, presenting various semisupervised K-means clustering techniques. Among these, literature [7] described a semisupervised K-means clustering algorithm based on genetics. e basic idea is to combine the unsupervised clustering accuracy (discrete degree) and the supervised clustering accuracy (supervised classi cation quality evaluation index) into a comprehensive semisupervised clustering quality evaluation index, use this index as the objective function, and optimize the clustering sample division using genetic algorithms [8]. Each chromosomal code corresponds to a set of grouping centers in the standard K-means technique throughout the optimization process.
To begin, when using a genetic algorithm, as a result of the clustering center encoded string disorder, two chromosomes between alleles and on the spatial location do not correspond to one another when the chromosome of the clustering center of mass is closer to the optimal; the system requires relatively local adjustment, crossover, and mutation calculations, of which the genetic algorithm is the most critical operation because it causes the position of the centroid to oscillate violently.
is paper offers PSO-K-means, a semisupervised K-means clustering technique based on dynamic particle swarm optimization, to solve the difficulties above. In this approach, the Euclidean space of all cluster samples is replicated using the dynamic particle swarm optimization search space, and each particle's position represents a cluster centroids group. e adaptive function is defined, and an iterative search is conducted to obtain a more accurate clustering result using the K-means technique. A new clustering quality evaluation function combining Euclidean distance and supervised information is constructed in light of the peculiarities of semisupervised clustering. e K-means technique is successfully applied to semisupervised clustering issues. e modified technique was evaluated on various standard UCI data sets and demonstrated superior clustering results [9]. e basic idea of the K-means clustering algorithm is to randomly select K initial clustering centers for the test object, calculate the distance between each seed clustering center and the center object, and then assign the object to the nearest center point. K-means clustering has become the most widely used algorithm among clustering algorithms due to its simplicity and high efficiency [10].

Background.
e scientific assessment of classroom teaching quality is a critical component of educational evaluation. To enhance the quality of instructors' classroom teaching, colleges and universities develop a classroom teaching quality assessment system and grade teachers' classroom teaching situations based on the evaluation index provided by the students in the class where the teacher teaches [11]. Usually, such evaluation activities are carried out at the end of the semester, and the Academic Affairs Office collects students' scores get the total teacher's classroom teaching score. A good evaluation system has a guiding function, which leads the evaluator to work hard towards a specific goal. At the same time, feedback is also provided for the evaluated teachers so that they can know their own teaching situation in time, revise teaching plans, and improve teaching methods according to the feedback information [12,13]. As the above evaluation system is carried out for teachers of all courses in the whole school (college), the evaluation indicators are single and stereotyped. e teaching quality cannot be evaluated according to the teaching characteristics of each discipline, so the impact of the teaching evaluation results on teaching is not apparent [14].
In today's society, colleges and universities provide a variety of activities, including teaching, scientific research, and development initiatives. e first and most basic duty, however, is talent development. As a teacher, the most common and most significant work should also be teaching. For most instructors, teaching and educating others is their primary obligation [15]. As a result, evaluating instructors' teaching abilities is a fundamental component of teacher assessment at colleges and universities. Teachers' teaching levels should be evaluated regularly to diagnose and improve teaching because college and university teaching is dynamic rather than static. Regular evaluation of teachers' daily teaching work has been widely used. Compared with the traditional teaching inspection, such as mutual lectures and observation of teaching, organized by the teaching management department, the evaluation of teachers' teaching level has the characteristics of collecting a large amount of information and checking specific items so that it can be a more comprehensive and objective diagnosis of the strengths and weaknesses in teaching. Suppose the obtained data can be stored in the computer, tracking analysis and research, and timely feedback to teachers and teaching management units. In that case, it will be more able to make a systematic analysis of the teaching situation from the statistical point of view to take significant teaching reform measures based on science [16].
Classroom instruction is the primary means by which a school achieves its educational objectives. From the standpoint of school administration, evaluating instructors' classroom teaching quality enables school leaders and managers to comprehend the achievement of teaching goals, completely grasp the teaching job, and improve teaching quality. Because the quality of teaching work substantially impacts the quality of trained talents, evaluating the quality of school teaching work has become an integral part of teaching management. Evaluating the quality of school teaching is difficult; the connection between instructors and students is nuanced during the teaching process, and various factors affect teaching quality. A critical issue is how to develop a scientific and accepted approach for evaluating teachers' teaching quality in a way that is objective and fair. ere appear to be some problems that some university teaching evaluation systems generally have: (1) e evaluation content is monotonous and not rich enough, and the evaluation subjects are relatively fixed and monotonous, lacking diversity and interactivity.
(2) e evaluation methods are single and not comprehensive enough. Some existing evaluation methods mostly do simple arithmetic, and statistical paper information lacks rationality and scientificity. e required evaluation time is long and increases the work burden of statisticians. ese will seriously affect the evaluation results.
(3) e evaluation feedback is not timely, and scores or grades only determine the evaluation results; however, such evaluation means cannot guide teachers' teaching work and students' learning well. Due to the rigid guidance of the baton of the college entrance examination, teachers tend to focus only on students' operational skills and basic knowledge in the evaluation process, ignoring students' emotions and values, which is not conducive to the overall development required of students in the context of quality education.

Our Contribution.
In response to the background mentioned above and methodological shortcomings, this paper uses a typical K-means model to find out the relationship between the indicators in the model and teachers' teaching ability, to identify the specific factors affecting teaching activities, and to guide the implementation of teachers' teaching work. We have the following contributions: (1) Using the cluster analysis method to properly analyze students' performance and teachers' evaluation, we can find out the common characteristics of classmates, discover their performance in certain courses, and analyze the learning characteristics of these courses to organize teaching better and accomplish teaching objectives. (2) e K-means method is used to process the data quickly, and the accuracy of the results fully meets the application conditions. (3) A large number of data experiments have proved that our method has a solid guiding effect on teachers' teaching quality and students' learning and helps students' learning progress.

College English Classroom Teaching
Quality Evaluation e weight of each assessment index and the construction of the evaluation object index set are connected to whether the evaluation is objective. e entire and complicated teaching process should be broken into multiple specific skill indexes when defining the evaluation index set. en, one or more evaluation indexes for in-depth and detailed assessment should be developed. is research creates the evaluation index set [5,6] based on the assessment's index methodology. e evaluation object index set and the weight of each evaluation index are developed based on this index system, as indicated in Table 1.
Taking the teaching quality evaluation of a certain college English class as an example, 5 students evaluated a certain English major teacher according to the evaluation indicators proposed in this paper (Table 1). e following takes "classroom teaching level" as an example to illustrate the fuzzy comprehensive evaluation process. e fuzzy set of evaluation results V � (A, B, C, D) corresponds to excellent, good, general, and poor, respectively.
e score of five students on this index is shown in

English Teaching Ability
Assessment. An information flow model with differential equations: where ω n is the measurement function of evaluation error and h is a multivariate value function of English teaching competence evaluation. Calculating the solution vector of English teaching ability assessment using the correlation fusion technique, which meets the following constraints, yields the feature training subset Si(I � 1,2, . . ., L) of teaching ability assessment: e data information flow model of English teaching ability assessment is created based on the preceding statistical measurement values for X (n): English teaching ability assessment has convergence solution and constraint conditions: According to the constructed data information flow model, a collection of scalar sampling sequence components is developed to establish a huge data distribution model, which offers an accurate data input foundation.

Quantitative Recursive Analysis of Teaching Ability
Assessment. To evaluate the significant ability assessment, a quantitative recursive analysis approach was used [4], and the control goal function of English teaching ability prediction and estimate was constructed as follows: max a∈A b∈B d∈D p∈P e probability density function can be obtained under the condition that the initial value of disturbance characteristics is fixed: Mathematical Problems in Engineering u c (t) � Kx c (t). (6) e function of the English teaching ability prediction and estimation system model is u: I × IR d ⟶ IR; after k−1 iteration, k > 1. e grey order of English teaching ability assessment satisfies N(k) < L. e quantitative recursive analysis approach is used: e big data clustering objective function is rough quantitative analysis x n N n�1 in the study, the following are the extraction findings of quantitative recursive aspects of teaching capacity assessment: where a 0 is the sampling amplitude of the initial English teaching ability assessment and b j is the oscillation decline value of English teaching ability assessment.

Optimization and Realization of English Teaching Ability Assessment
Model. e analysis of the English teaching ability of the big data information model is based on the quantitative appraisal ability.

z(t) � x(t) + iy(t) � a(t)e iθ(t) + n(t).
(10) X′(k) was obtained by amplitude randomization of English teaching ability using the substitution data method. us, the utilization ratio could be expressed as follows: e clustering assessment is accomplished using the linear correlation feature fusion technique, and the output expression is as follows: e big data stream X(I) is divided into P(I) submatrix x ij with the size of n ij × m.

K-Means Algorithm
(1) Select K initial centers of mass (K needs to be specified by the user), the initial centers of mass can be chosen randomly, and each center of mass is a class.

Simulation Experiment Analysis
e performance of big data analysis for assessing English teaching skills is tested using the MATLAB simulation analysis approach. Data sampling is done using a statistical analysis approach to assess English teaching competence. Dx 2 is the teaching ability assessment decision threshold value. We collect commonly used teaching quality evaluation datasets and divide the rst 75 percent of the dataset into training sets and the remaining 25 percent as test sets. Table 3 provides the assessment accuracy and other indicators test results. According to the ndings, the approach used in this study has higher accuracy in evaluating teaching skills and a higher utilization rate of teaching resources. e whole K-means algorithm is based on the EM algorithm. It is the E step to mark sample points based on the expectation of the center point and the M step to move the center point to the center of the new standard sample point. e application of the EM algorithm is far from limited to K-means. It also plays an essential role in the Gaussian mixture model, and GMM is trained with the GEM algorithm (the promotion of the EM algorithm). e core of the EM algorithm is to divide variables to be optimized into hidden variables, parameters, and observable variables [7,8].
e comparative analysis of algorithm improvement results is shown in Figure 1.   en, calculate the distance of each data point to these centers separately and use the one with the shortest distance as its category or center point. In this way, each point will have a center point, resulting from randomly generated center points. It can be seen that the clustering is not accurate, as shown in Figure 2.
e improved algorithm is used to converge the results of the last clustering and recalculate the mean value as the new clustering center and recluster the center points according to the clustering center points. e result is shown in Figure 3.
Conclusion: in general, the K-means clustering algorithm randomly selects K sample sets waiting for clustering, and the completion of clustering is directly related to the selection of the initial clustering center. Once the sample selection result is unreasonable, the complexity of the operation process will be increased accordingly, which will mislead the whole clustering process and nally get no good result. By adjusting the number and mean of initial classes needed by the modi ed K-means clustering algorithm's initial clustering center, the e ciency of clustering and classi cation accuracy may be increased.
Randomly set 10% of all data as existing label data and instance limit data as monitoring information. To eliminate the randomness of setting supervision information, ve groups of di erent supervision information sets were generated for each experimental data set. For each group of the supervision information set, each algorithm was run 10 times to calculate the average clustering accuracy. To facilitate readers to observe the experimental results, we give the radar diagram of the results in Table 4, as shown in Figure 4. e following can be seen from the clustering results on multiple data sets in Table 4: (A) Compared with the traditional K-means algorithm, the clustering accuracy of the improved PSO-Kmeans algorithm has been greatly improved due to centroid optimization. (B) Compared with Ga-K-means, due to the higher coupling degree between particle swarm space and Euclidean space, semisupervised information is added in sample classi cation, and the clustering result PSO-K-means is generally better. e anomaly in the IRIS data set is due to the high clustering accuracy of the Iris data set and itself, which fails to re ect the algorithm's advantages.
Compared with the PSO-K-means algorithm of static and dynamic population management, although the population number of dynamic PSO is always less than or equal to that of static PSO in the operation process, it is due to the updating function of dynamic population management that continuously adds new particles with the heuristic ability and deletes particles without heuristic ability. Figure 5 tests the clustering accuracy of various algorithms when label data accounts for di erent percentages (5%-40%) of the data set. As can be seen from the results, when the supervised information is small, there is no signi cant di erence in clustering result precision between the two algorithms. When the percentage of supervised data gradually increases, the clustering accuracy of Ga-K-means no longer increases, and the gap with PSO-K-means becomes larger and larger. is shows that a new neighborhood measure calculation function with supervisory information is used in the clustering process. Compared with simple Euclidean distance, it can better use monitoring information and get better clustering accuracy. e data from the rst 1 to 10 groups in Figure 6 is used as training data (containing teacher data), while the data from the remaining 4 groups is utilized as veri cation data (excluding teacher data) to check the trained neural network model's prediction results.  is research uses the teacher supervision group's assessment value after the class as sample data to train the network. Its relevance stems from the fact that, although students ll out instructor indications, the nal assessment results represent the evaluation thoughts of the supervision group's specialists.

Conclusion
In order to continuously improve the teaching level and organizational management level of foreign language teachers, attention should be paid to the assessment of foreign language teachers, the evaluation of teachers' teaching level should be included in the regular management plan, the collection of assessment information should be institutionalized and standardized, and e orts should be made to improve the reliability and validity of evaluation. e work is recognized by both teaching parties. We should encourage the excellent ones and promote the underachievers so that the level of college English teaching can reach a new level. Firstly, the method minimizes the in uence of external and internal interference factors on evaluation results. It solves the problem of multi-index comprehensive evaluation and fully considers the relationship between each index and the in uence of individual factors in evaluation. Secondly, compared with the traditional evaluation methods, the application of fuzzy comprehensive evaluation can not only objectively evaluate the class teaching quality in the grades of an excellent, good, medium, and poor but also avoid the consequences of excessive fuzziness and subjectivity in evaluation and reduce the interference of fuzzy factors in evaluation to a certain extent. Make the evaluation result more scienti c and reasonable. As a result, the application of fuzzy comprehensive evaluation in college English classroom teaching quality assessment gives a new way of evaluating English classroom teaching, and the approach is simple to use, to some extent, to overcome the evaluation's current issues.  Although this thesis has achieved specific research results and has some significance in evaluating classroom teaching quality in colleges and universities, some shortcomings still need further improvement and refinement. Future research will be conducted in the following two aspects.
(1) In the research on classroom teaching quality evaluation, there is no clear standard for the system and evaluation assessment subject. At the same time, this paper builds the evaluation index system based on relevant literature and the current evaluation system. Its rigor and extensiveness need to continue to be tested, so in future research, we should try to adopt specific methods to conduct experimental research on the evaluation index system, and we must come up with accurate and appropriate results to build a more scientific and complete teaching quality evaluation index system. (2) e results of the algorithm currently used are not good enough, and it is the defect of the algorithm itself; although the processing speed is fast, it is easy to appear outlier points. I hope that the subsequent research can both meet the requirements of speed and improve the accuracy rate of the classification.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
e author declares no conflicts of interest.