Student learning state classification method based on campus global human body information perception

Abstract

The invention specifically relates to a student learning state classification method based on campus global human body information perception, and belongs to the technical field of artificial intelligence education data mining. According to the classification method, firstly, a built-in body temperature sensor of a student smart bracelet is used for obtaining body temperature and wrist temperature information, a heart rate sensor is used for obtaining heart rate information, a blood pressure sensor is used for obtaining diastolic pressure and systolic pressure information, the data are transmitted into a server through an API interface, and the information data of the body temperature, the wrist temperature, the heart rate, the diastolic pressure and the systolic pressure are decoded and stored. Secondly, through data visualization, arranging the data into a scatter diagram form, checking and removing abnormal data, analyzing linear correlation between features, and eliminating redundant features; next, feature extraction is carried out in a statistical feature, K-means clustering and frequent set discovery mode, and preprocessing is carried out through a normalization method; then, three machine learning algorithms of KNN, random forest and XGBoost are selected to carry out training and preliminary recognition on the normalized data set, and finally, a preliminary recognition result and the data set are combined into a new data set to realize student learning state classification based on campus global human body information perception through a deep knowledge tracking model.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence education data mining, and specifically relates to a student learning status classification method based on campus-wide human body information perception. Background Art

[0002] The application of artificial intelligence technology in education enables teachers to perform management functions more effectively, greatly reducing the workload when performing administrative functions, so that teachers can focus on teaching quality and improve teaching efficiency. As the main body of teachers' teaching activities, students' behavior status is a direct reflection of teaching activities. Teachers use artificial intelligence technology to monitor students' learning behavior status, accurately analyze and grasp students' behavior patterns to intervene in time, and dynamically adjust the teaching progress according to students' learning behavior status to improve classroom efficiency. Most researchers usually use cameras and microphones to collect images and voice information of students during class when monitoring students' classroom learning status, but this method is only suitable for indoor teaching environments. In order to reduce noise interference, high-precision acquisition equipment is also required. Some researchers study the relationship between physiological information and student behavior activities by monitoring physiological information, but this method has the problem of single data. At the same time, there will be subjective intervention when the experimenter is conscious, which will lead to the problem of untrue experimental data.

[0003] In addition, previous studies have found that in addition to the students' learning behavior, sleep and entertainment behavior also affect students' learning efficiency. Sleep is an active, repetitive behavior that plays an important role in the learning process and memory consolidation. Lack of sleep or too much sleep will have a certain negative impact on learning and memory. Entertainment, including watching movies or TV series, reading novels, and playing games, has a certain positive effect on training students' visual-spatial skills, and can also promote students' reflection and creativity.

[0004] Therefore, in order to assist teachers in teaching, accurately monitor students' learning status and improve learning efficiency, there is an urgent need for a method that can solve problems such as environmental limitations, expensive equipment, single data and manual intervention to classify students' three learning states: learning, entertainment and sleep. Summary of the invention

[0005] In view of the above-mentioned defects existing in the prior art, the present invention provides a method for classifying student learning status based on campus global human information perception. The method first uses the built-in temperature sensor of the student smart bracelet to obtain body temperature and wrist temperature information, the heart rate sensor to obtain heart rate information, and the blood pressure sensor to obtain diastolic and systolic blood pressure information. These data are transmitted to the server through the API interface, and the body temperature, wrist temperature, heart rate, diastolic blood pressure, and systolic blood pressure information data are decoded and saved. Secondly, through data visualization, the data is organized into a scatter plot, abnormal data is checked and removed, the linear correlation between features is analyzed, and redundant features are eliminated. Next, feature extraction is performed using statistical features, K-means clustering, and frequent set discovery, and normalization methods are used for preprocessing. Then, three machine learning algorithms, KNN, random forest, and XGBoost, are selected to train and preliminarily identify the normalized data set. Finally, the preliminary recognition results and the data set are combined into a new data set to achieve the classification of student learning status based on campus global human information perception through a deep knowledge tracking model.

[0006] The present invention is achieved through the following technical solutions:

[0007] A method for classifying student learning status based on campus-wide human body information perception specifically includes the following steps:

[0008] Step 1: Data collection:

[0009] Before data collection, students need to undergo physical examinations and fill out informed consent forms. Students wear wristbands and move around the campus as normal, while recording their time in three behavioral states at school. The wristband sends physiological information data to the server for reception, and at the same time, the data is marked with behavioral status;

[0010] Step 2: Data visualization:

[0011] For the labeled data set, the cluster scatter plot is used to observe the data distribution, the heat map is used to calculate the linear correlation between features, and the feature relationship diagram is used to roughly observe the connection between data and behavior status, so as to better remove outliers and redundant features and extract useful information;

[0012] Step 3: Data preprocessing:

[0013] After visual analysis, the program traverses the data set to remove abnormal body temperature data and wrist temperature, heart rate, diastolic blood pressure, and systolic blood pressure data within the corresponding time, while only retaining features with small linear correlation. Statistical features, K-means clustering algorithm and Apriori frequent set discovery algorithm are used to extract features from the data set, and finally normalized;

[0014] Step 4: Build a machine learning model:

[0015] The processed data sets were trained and preliminarily identified using three machine learning models: KNN, random forest, and SVM. The parameters of each machine learning model were adjusted to find the optimal threshold, and the recognition results of the three models were added to the normalized data set.

[0016] Step 5: Classification of student learning status:

[0017] The new data set is used as the input of the deep knowledge tracking model, and the parameters of the deep knowledge tracking model are adjusted to optimize the model effect. Finally, the final student learning status classification is performed.

[0018] Furthermore, in step 1, the wristband worn by the student has a built-in temperature sensor, a heart rate sensor, and a blood pressure sensor; the temperature sensor is used to obtain body temperature and wrist temperature information, the heart rate sensor is used to obtain heart rate information, and the blood pressure sensor is used to obtain diastolic and systolic blood pressure information. Students use the app to clock in and record the start and end time of their three behavioral states: study, entertainment, and sleep. If the student forgets to clock in or is unclear about the time period of the behavioral state, the physiological information data of this time period will be ignored.

[0019] Furthermore, in step 1, various sensors built into the wristband send physiological information data to the server through the NB-IOT network using the API interface, and then parse and save the obtained data into a csv file. At the same time, in order to ensure sufficient usage time and physiological information data, the relationship between the power consumption of the wristband and the transmission frequency must also be considered.

[0020] Furthermore, the data set in step 2 includes parsed heart rate, wrist temperature, heart rate, diastolic blood pressure, systolic blood pressure data, and corresponding behavioral state labels.

[0021] Furthermore, in step 3, feature extraction is performed based on the K-means clustering algorithm, which specifically includes the following:

[0022] A. Delete the labels from the data set in step 2 to form a new data set as the input of the K-means clustering algorithm for classification. Use the profile analysis method to evaluate the model and find the most appropriate K value. Use the classified cluster results as new features.

[0023] B. Each cluster has a centroid, and the Euclidean distance from each sample point to the centroid of the category is used as a new feature.

[0024] Furthermore, in step 3, frequent set discovery is performed based on the Apriori algorithm, which specifically includes the following:

[0025] A. Quantify the range of each feature value of the data set in step 2 into intervals;

[0026] B. Through the Apriori algorithm, the physiological information of students and the support of each student are mined;

[0027] The support calculation adopts the following formula:

[0028]

[0029] The support refers to the probability that event A and event B appear simultaneously in the total event I.

[0030] C. The support of the values ​​in the corresponding range of the five physiological information is used as a new feature.

[0031] Furthermore, in step three, feature extraction is performed based on statistical features, the width and step size of the sliding window are set, and the mean, mode, median, variance, minimum and maximum values ​​of the physiological information data in step two are calculated in each window as new features.

[0032] Furthermore, the data normalized in step 3 include new features extracted based on the K-means clustering algorithm, the Apriori frequent set discovery algorithm, and statistical features;

[0033] The normalization process uses the following formula:

[0034]

[0035] Among them, x is the data value, max is the maximum value, and min is the minimum value;

[0036] Compared with the prior art, the advantages of the present invention are as follows:

[0037] The present invention discloses a method for identifying student learning behavior states based on campus-wide human body information perception. The method collects students' body temperature, wrist temperature, heart rate, diastolic blood pressure, systolic blood pressure and other physiological information data through a wristband, and analyzes students' three behavioral states of learning, entertainment and sleeping. Compared with other collection methods, the method solves the problems of environmental limitations, expensive equipment, single data and manual intervention, making teacher classroom intervention and improvement of teaching methods more efficient. In addition, it is found through visualization that there are too few features to be well classified, and then three methods, namely statistical features, K-means clustering and frequent set discovery, are used for further feature extraction. Finally, the preliminary recognition results of the machine learning model are used as the features in the deep knowledge tracking model, which combines the advantages of multiple algorithms and increases the practicability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0039] Figure 1 It is a flow chart of a method for identifying student learning behavior status based on campus-wide human body information perception according to the present invention;

[0040] Figure 2 It is a schematic diagram of data collection using a wristband according to the present invention;

[0041] Figure 3 The present invention draws cluster scatter plots, heat maps, and feature relationship diagrams through data visualization analysis, where (a) is the distribution of the data cluster scatter plot after behavior labeling, (b) is the linear correlation between students' physiological information features, and (c) is a feature relationship diagram, which simply classifies the three learning states of study, entertainment, and sleep.

[0042] Figure 4 It is a schematic diagram of feature extraction through statistical features in the present invention.

[0043] Figure 5 The present invention uses K-means clustering algorithm for feature extraction, and the silhouette coefficient distribution under different k values.

[0044] Figure 6 The present invention uses the Apriori frequent set discovery algorithm for feature extraction, and the results of mining association rules of physiological information of five students are obtained.

[0045] Figure 7 This is a flowchart of the final recognition and classification framework using machine learning algorithms and deep knowledge tracking algorithms of the present invention. DETAILED DESCRIPTION

[0046] In order to clearly and completely describe the technical solution and its specific working process of the present invention, the specific implementation methods of the present invention are as follows in conjunction with the accompanying drawings of the specification:

[0047] Example 1

[0048] like Figure 1 FIG. 1 is a flow chart of a method for identifying a student learning behavior state based on campus global human body information perception according to an embodiment of the present invention. The identification method specifically includes the following steps:

[0049] Step 1: Data collection:

[0050] This embodiment recruited 30 college students aged between 19 and 21, including 20 boys and 10 girls. Physical examination activities included internal medicine, surgery, ophthalmology, otolaryngology, electrocardiogram, DR chest radiograph, and blood routine. The physical examinations before data collection were all qualified, and no students showed any symptoms of disease during the data collection process. And each student filled out an informed consent form, the contents of which are mainly as follows: (1) research background and research purpose; (2) data collection time is from October 1, 2022 to November 30, 2022; (3) confidentiality of personal information, only for scientific research, no information will be leaked; (4) you can exit the ongoing data collection process at any time.

[0051] The wristband used in this embodiment is the B2315+ series wristband developed by Shanghai Oufu Communication Technology Co., Ltd. The built-in body temperature sensor of the wristband obtains body temperature and wrist temperature information, the heart rate sensor obtains heart rate information, and the blood pressure sensor obtains diastolic and systolic blood pressure information. These data information are sent every five minutes through China Telecom's NBIOT Internet of Things card inside the wristband, and reported to China Telecom AEP platform using CoAP protocol. Then, the local server uses the web application framework flask to create an API interface, and the data is saved in the local server in the form of a string. Finally, the data is decoded, classified, and exported for more convenient data mining and learning behavior analysis. The data collection framework is as follows: Figure 2 shown.

[0052] The data directly obtained through the API interface is shown in Table 1; the attribute information is as follows: (1) bodyTemperature: body temperature; (2) wristTemperature: wrist temperature; (3) BTUtcTime, time, year, month, day, hour, minute, second; (4) diastolic: diastolic pressure; (5) shrink: systolic pressure; (6) heartbeat: heart rate; (7) IMEI, bracelet ID, which is also the ID of each student; (8) type, data type, for example, type equals 14, which means that the data is body temperature and wrist temperature information, type equals 8, which means that the data is diastolic pressure and systolic pressure information, and type equals 6, which means that the information is heart rate information.

[0053] Table 1 Part of the data directly obtained by the API interface

[0054]

[0055] Furthermore, the data in Table 1 were decoded using a python program, divided into three types of files: "blood pressure information", "body temperature information", and "heart rate information", and exported. Finally, based on the way students clocked in through the app to record the start and end time of their three behavioral states of study, entertainment, and sleep, the blood pressure, body temperature, and heart rate information of the students during class time were selected and organized into a physiological information data set. Among them, if a student forgets to clock in or the student is unclear about the time period of the behavioral state, the physiological information data of this time period will be ignored. In addition, since there are differences in the total length of time of the three behavioral states collected, the physiological information data of the behavioral state with the smallest total length is used as the main data, and the physiological information data of the redundant time periods of the other two behavioral states are deleted. The summary table of some data sets of physiological information is shown in Table 2;

[0056] Table 2. Partial data set arrangement of physiological information

[0057] BodyTem WristTem Hr dio shrink type 36 29.3 79 73 112 sleep 36.1 30.1 92 76 118 sleep 36.2 30.9 73 75 115 sleep 36.5 32.6 81 69 111 study 36.5 32.5 81 77 117 study 36.4 32.1 76 77 115 study 36.4 32.2 84 81 120 yule 36.9 35 71 73 113 yule 37 35.5 70 70 111 yule

[0058] Step 2: Data visualization:

[0059] The relationship between various physiological information data and student behavior status is visualized, and the data is organized into a more understandable form. By using cluster scatter plots, heat maps, and feature relationship diagrams, a series of experimental steps such as removing outliers and redundant features and feature extraction are determined.

[0060] Figure 3 In (a), overall, the horizontal axis represents the five features of body temperature, wrist temperature, heart rate, diastolic blood pressure, and systolic blood pressure (from left to right), and the vertical axis represents the three behavioral states of entertainment, learning, and sleeping (from top to bottom). Each sub-graph represents the numerical distribution of different features under different behavioral states, with the horizontal axis representing the value range of each feature and the vertical axis representing each behavioral state. Figure 3 In (b), the linear correlation between each feature can be known based on the depth of the color and the calculated Pearson correlation coefficient. In order to avoid feature redundancy, some features can be deleted instead of using five features at the same time. Figure 3 In (c), the diagonal line is the histogram of each physiological information, and the off-diagonal line is the correlation diagram between two different physiological information.

[0061] Step 3: Data preprocessing:

[0062] According to the data visualization analysis, some data with body temperature below 35 degrees Celsius were found. Since the data came from healthy students, the program traversed the data set to remove these abnormal body temperature data and wrist temperature, heart rate, diastolic blood pressure, and systolic blood pressure data within the corresponding time, thereby reducing noise interference. At the same time, since body temperature and wrist temperature, diastolic blood pressure and systolic blood pressure have a very high linear correlation, only three types of features, body temperature, heart rate, and diastolic blood pressure, were retained in the end.

[0063] Furthermore, based on statistical features, K-means clustering algorithm and Apriori frequent set discovery algorithm, features of the data set are extracted. Figure 4 In , the width of the sliding window is set to 10 and the step size is 2. In each window, the mean, mode, median, variance, minimum and maximum value of each feature in the data set after removing abnormal data and redundant features are calculated, and these are used as the extracted features.

[0064] Figure 5 In the above example, the dataset after removing abnormal data and redundant features is delabeled as the input of K-means clustering algorithm, and the model is evaluated by the contour analysis method to find the appropriate k value. The contour analysis method measures the degree of clustering of samples in the cluster. The calculation formula of the contour coefficient S is as follows:

[0065]

[0066] Among them, a is the average distance between each sample data and other points in the cluster, and b is the average distance between each sample data and all points in the nearest cluster.

[0067] The k value corresponding to the largest silhouette coefficient is the optimal k value, that is, the data can be preliminarily divided into two clusters according to the similarity of the data. The classified cluster results 0 and 1 and the Euclidean distance from each sample point to the final centroid of the category are used as the extracted features.

[0068] Due to differences in environment and living habits, each student's physical fitness is different. Figure 6 In this paper, the Apriori algorithm is used to mine association rules for each student's physiological information data, and the frequent sets of each student's body temperature, wrist temperature, heart rate, diastolic blood pressure and systolic blood pressure are found. The data set after removing abnormal data and redundant features is delabeled as the input of the Apriori algorithm, and the Apriori algorithm is used to mine association rules. The numerical range of each feature in the student learning behavior status classification data set is quantified and intervalized, and then the support of the values ​​in the corresponding range of the five physiological information is used as the extracted features.

[0069] Furthermore, the features extracted above are respectively subjected to maximum-minimum normalization processing to eliminate the influence of dimension;

[0070] The maximum value max and the minimum value min in the data column are used for standardization. The standardized value is between [0,1]. The calculation method is to subtract the data x from the minimum value min of the column and then divide it by the range, that is:

[0071]

[0072] Step 4: Build a machine learning model:

[0073] The data preprocessed in step 3 are used to train and identify the learning behavior status of students by using three machine learning algorithms: KNN, random forest, and SVM to extract the feature sets;

[0074] KNN classifies objects by searching for the K most similar instances, i.e., K neighbors, in the entire training set and assigning a common output variable to all these K instances.

[0075] During the training process of Random Forest, each decision tree is constructed based on a bootstrap sample of the training set. During the classification process, the decision for the input instance is made based on majority voting;

[0076] SVM plots the data items as points in n-dimensional space, where n is the number of input features. Based on this, the support vector machine finds an optimal boundary, called a hyperplane, which best separates the possible outputs by class labels. The distance between the hyperplane and the nearest class point is called the margin. The optimal hyperplane has the largest margin and can classify the points so that the distance between the nearest data point and the two classes is maximized.

[0077] Step 5: Classification of student learning status:

[0078] like Figure 7 As shown in the figure, the machine learning classification results in step 4 and the new features extracted in step 3 are combined into a new feature set as the input of the deep knowledge tracking model to perform the final student learning status classification, as follows:

[0079] Deep knowledge tracking technology estimates the learner's internal, implicit knowledge state or mastery level based on the student's external, explicit learning performance or behavior sequence. Taking feature fusion as the starting point, the assessment behavior and knowledge structure features are integrated into the deep knowledge tracking model, and a deep knowledge tracking model integrating domain features is constructed, in order to further improve the learner's knowledge tracking effect.

[0080] The classification method described in this embodiment was applied to recruit 30 college students aged between 19 and 21 years old in a certain university. Among them, 20 boys and 10 girls were used for experiments. 400,000 pieces of physiological information data were collected, and the data were classified and sorted by data cleaning; then the discrete data and five original physiological characteristics were sorted and analyzed for correlation using visual data mining technology, and the next step of data preprocessing including removal of abnormal points and redundant features, feature extraction, and normalization was determined. Among them, feature extraction adopted three methods: statistical features, K-means clustering, and frequent set discovery; then the parameters of KNN, random forest, and SVM classifiers were adjusted to achieve the optimal solution, and the data set was trained and preliminarily identified. Finally, the deep knowledge tracking algorithm was used to classify the final student learning status. The experimental results show that the model can well identify the three behavioral states of students' learning, entertainment, and sleeping at any time and any place. Under the joint action of the three feature extraction methods, the model can be improved by 39.345%, and the accuracy rate reached 97.689%. Among them, the feature extraction effect using the frequent set discovery method is the most obvious. In this experiment, a new student learning status recognition dataset was provided. Visual data mining analysis was used to determine the experimental steps. K-means clustering and Apriori algorithm were tried for feature extraction. It was found that the Apriori algorithm can reduce the dimension of the dataset but have a higher impact effect. In addition, the results of the original feature correlation analysis using visual data mining were used to eliminate two types of features with high correlation, and the discovery effect was further improved by 0.116.

[0081] The preferred embodiments of the present invention are described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, a variety of simple modifications can be made to the technical solution of the present invention, and these simple modifications all belong to the protection scope of the present invention.

[0082] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations.

[0083] In addition, various embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the present invention, they should also be regarded as the contents disclosed by the present invention.

Claims

1. A method for classifying student learning status based on campus-wide human body information perception, characterized in that: The specific steps include: Step 1: Data collection: Before data collection, students need to undergo physical examinations and fill out informed consent forms. Students wear wristbands and move around the campus as normal, while recording their time in three behavioral states at school. The wristband sends physiological information data to the server for reception, and at the same time, the data is marked with behavioral status; Step 2: Data visualization: For the labeled data set, the cluster scatter plot is used to observe the data distribution, the heat map is used to calculate the linear correlation between features, and the feature relationship diagram is used to roughly observe the connection between data and behavior status, so as to better remove outliers and redundant features and extract useful information; Step 3: Data preprocessing: After visual analysis, the program traverses the data set to remove abnormal body temperature data and wrist temperature, heart rate, diastolic blood pressure, and systolic blood pressure data within the corresponding time, while only retaining features with small linear correlation. Statistical features, K-means clustering algorithm and Apriori frequent set discovery algorithm are used to extract features from the data set, and finally normalized; Step 4: Build a machine learning model: The processed data sets were trained and preliminarily identified using three machine learning models: KNN, random forest, and SVM. The parameters of each machine learning model were adjusted to find the optimal threshold, and the recognition results of the three models were added to the normalized data set. Step 5: Classification of student learning status: The new data set is used as the input of the deep knowledge tracking model, and the parameters of the deep knowledge tracking model are adjusted to optimize the model effect. Finally, the final student learning status classification is performed.

2. A method for classifying student learning status based on campus-wide human body information perception as claimed in claim 1, characterized in that: In step 1, the bracelet worn by the students has a built-in temperature sensor, heart rate sensor, and blood pressure sensor; the temperature sensor is used to obtain body temperature and wrist temperature information, the heart rate sensor is used to obtain heart rate information, and the blood pressure sensor is used to obtain diastolic and systolic blood pressure information. Students use the app to punch in to record the start and end time of their three behavioral states: study, entertainment, and sleep. If the student forgets to punch in or is unclear about the time period of the behavioral state, the physiological information data of this time period will be ignored.

3. A method for classifying student learning status based on campus-wide human body information perception as claimed in claim 1, characterized in that: In step 1, the various sensors built into the bracelet use the API interface to send physiological information data to the server through the NB-IOT network, and then parse and save the obtained data into a csv file. At the same time, in order to ensure sufficient usage time and physiological information data, the relationship between the power consumption of the bracelet and the transmission frequency must also be considered.

4. The method for classifying student learning status based on campus-wide human body information perception as claimed in claim 1, characterized in that: The data set in step 2 includes the parsed heart rate, wrist temperature, heart rate, diastolic blood pressure, systolic blood pressure data, and the corresponding behavioral state labels.

5. The method for classifying student learning status based on campus-wide human body information perception as claimed in claim 1, characterized in that: In step 3, feature extraction is performed based on the K-means clustering algorithm, which includes the following: A. Delete the labels from the data set in step 2 to form a new data set as the input of the K-means clustering algorithm for classification. Use the profile analysis method to evaluate the model and find the most appropriate K value. Use the classified cluster results as new features. B. Each cluster has a centroid, and the Euclidean distance from each sample point to the centroid of the category is used as a new feature.

6. The method for classifying student learning status based on campus-wide human body information perception as claimed in claim 1, characterized in that: In step 3, frequent set discovery is performed based on the Apriori algorithm, which includes the following: A. Quantify the attribute interval of each feature value range of the data set in step 2; B. Through the Apriori algorithm, the physiological information of students and the support of each student are mined; The support calculation adopts the following formula: Among them, support refers to the probability that event A and event B appear simultaneously in the total event I. C. The support of the values ​​in the corresponding range of the five physiological information is used as a new feature.

7. The method for classifying student learning status based on campus-wide human body information perception as claimed in claim 1, characterized in that: In step three, feature extraction is performed based on statistical features, the width and step size of the sliding window are set, and the mean, mode, median, variance, minimum and maximum values ​​of the physiological information data in step two are calculated in each window as new features.

8. The method for classifying student learning status based on campus-wide human body information perception as claimed in claim 1, characterized in that: The data normalized in step 3 include new features extracted based on K-means clustering algorithm, Apriori frequent set discovery algorithm and statistical features; The normalization process uses the following formula: Among them, x is the data value, max is the maximum value, and min is the minimum value.

Images