A college student in economic difficulties recognition method and system based on EW-KMeans algorithm
By calculating the feature index weights using the EW-KMeans algorithm and the entropy weight method, and combining them with the CART decision tree for the identification of economically disadvantaged students in colleges and universities, the problems of misjudgment and omission in traditional methods are solved, and the identification effect with higher accuracy and interpretability is achieved.
Patent Information
- Application Number
- CN202610895942.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-22
- Publication Date
- 2026-11-06
AI Technical Summary
Existing technologies for identifying economically disadvantaged students in colleges and universities suffer from misjudgments, omissions, and a lack of interpretability. In particular, traditional methods rely on a single threshold, leading to group bias and misjudgments based on low consumption, failing to identify hidden economically disadvantaged groups, and the output of the scoring model lacks interpretability of behavioral rules.
The EW-KMeans algorithm, combined with entropy weighting and CART decision tree, was used to identify economically disadvantaged students by collecting student consumption and library data, calculating feature index weights, performing clustering and rule mining.
It improves the accuracy and adaptability of identifying economically disadvantaged students, provides interpretable decision-making rules, reduces misjudgments and omissions, and achieves accurate identification and dynamic adjustment of hidden economically disadvantaged groups.
Smart Images
Figure CN123049186A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of identification technology for economically disadvantaged students in colleges and universities, specifically involving a method and system for identifying economically disadvantaged students in colleges and universities based on the EW-KMeans algorithm. Background Technology
[0002] Identifying economically disadvantaged students in universities is a fundamental step in student financial aid programs. Traditional methods rely primarily on students' submitted proof of family financial hardship, information on urban and rural minimum living allowances, and reviews by auditors. These methods suffer from problems such as information lag, high review costs, and susceptibility to false reporting or underreporting. In recent years, methods for identifying economically disadvantaged students based on campus card spending data have gained increasing attention. The basic idea is to compare indicators such as average daily spending and spending frequency with a unified threshold across the entire university; students whose spending is significantly below the threshold are considered potential economically disadvantaged.
[0003] However, existing technologies for identifying economic hardship based on consumer data have the following inherent drawbacks: (1) Single threshold leads to group bias: For groups with low participation in campus activities (such as going to the canteen less, going to the library less, and spending less on campus), their absolute value of campus consumption data is extremely low. Under the unified threshold of the whole school, almost all of this group is marked as "economically disadvantaged", resulting in a large number of invalid warnings.
[0004] (2) Some students’ low consumption is not due to economic hardship, but rather to “low-desire” lifestyle habits (such as dieting, simple living, and eating out). Traditional methods directly equate low consumption with economic hardship, resulting in the misjudgment of “false economic hardship”; while some students who frequently visit the library have simple and unpretentious consumption patterns, but are ignored by traditional models because they lack significant “low consumption” signals, resulting in the inability to identify the hidden economic hardship group.
[0005] (3) Lack of interpretability: Existing scoring models often output a continuous economic hardship score or a black box classification result, which cannot provide specific behavioral rules for "why the student was judged to be economically hardship", which is not conducive to review. Summary of the Invention
[0006] This invention addresses the shortcomings of existing technologies by providing a method and system for identifying economically disadvantaged students in colleges and universities based on the EW-KMeans algorithm. This method avoids misjudgments and omissions caused by a one-size-fits-all approach across the entire university, and has higher identification accuracy, stronger group adaptability, and better decision interpretability.
[0007] This invention provides the following technical solution: Firstly, this paper provides a method for identifying economically disadvantaged college students based on the EW-KMeans algorithm, including: Collect all students' consumption data and library access control data, and compile all characteristic indicators for each student, including consumption indicators and library indicators; Based on all the characteristic indicators of each student, the entropy weight method is used to calculate the weight of each characteristic indicator, and the entropy weight truncation and redistribution method is introduced to correct the calculated weight of each indicator. Based on the corrected weights of each indicator and the statistical characteristic indicators of each student, a weighted feature matrix is constructed for each student. After determining the optimal number of clusters K using the elbow method, all students are clustered based on the weighted feature matrices of all students to obtain several student groups. For each student group, the economic hardship score of each student in the group is calculated and classified based on the relative benchmark of their consumption indicators. Then, decision trees are used to mine rules within each group to identify all suspected economically disadvantaged students in each group.
[0008] Optionally, the consumption indicators include: number of transaction days, average daily consumption, single transaction amount, and breakfast rate; the library indicators include: library behavior presence characteristics and library behavior intensity; the library behavior presence characteristics and library behavior intensity are determined based on the number of times the current student accesses the library within the statistical period.
[0009] Optionally, the existence characteristic of the library behavior is: ; in, For the first Existential characteristics of students' library behavior. , For the number of students, For the first The number of times a student accesses the library during the statistical period; The intensity of the library behavior is: ; in, , For all The average value, For all The standard deviation.
[0010] Optionally, based on all the characteristic indicators of each student, the weight of each characteristic indicator is calculated using the entropy weight method, and the calculated weight of each indicator is corrected by introducing an entropy weight truncation and redistribution method, specifically as follows: Step D1: Calculate the initial weight of each feature indicator using the entropy weight method, and set the upper limit of the weight threshold based on the reciprocal of the number of feature indicators. ; Step D2: Determine whether the current weights of all feature indicators are less than the upper limit of the weight threshold. If yes, output the current weights of all feature metrics; otherwise, start from a value greater than the upper limit of the weight threshold. From all the current weights, select the current weight of any feature indicator, and adjust the current weights of some feature indicators according to the following formula; ; in, From greater than the upper limit of the weight threshold The first selected from The current weights of each feature indicator, For the revised first The current weights of each feature indicator, Less than the upper limit of the weight threshold The The current weights of each feature indicator, For the revised first The current weights of each feature indicator, This represents the sum of the excess amounts of all features that exceed the threshold. Less than the upper limit of the weight threshold The current number of weights; Step D3: Repeat step D2 until the current weights of all feature indicators are less than the upper limit of the weight threshold. .
[0011] Optionally, the weighted feature matrix of each student is represented as: ; in, For the first The weighted feature matrix of each student For the first The first student Each weighted feature indicator For the first The weights of each feature indicator after correction For the first The first student Each characteristic indicator.
[0012] Optionally, the calculation of economic hardship scores for each student within the group based on a relative benchmark of their consumption indicators and the subsequent classification into levels specifically involves: For each student group, calculate the 75th percentile of all their consumption indicators, and use the 75th percentile as the benchmark value for the current student group in terms of current consumption indicators. Calculate each student's individual economic hardship score according to the following formula; ; in, For the first The student in the first The one-way economic difficulty score of each consumption indicator, For the first The first student A weighted consumption indicator, For the first The first among the student groups The benchmark value for each consumption indicator; By weighting and aggregating all individual economic hardship scores, a student's economic hardship score is obtained: ; in, For the first Each student's score on their level of financial hardship For the quantity of consumption indicators, for The fusion weight; Each student is classified into different levels of financial hardship based on their financial hardship score.
[0013] Optionally, the step of using decision trees to perform rule mining within each group to identify all suspected economically disadvantaged students in each group specifically involves: For each group, a CART decision tree is trained using the weighted feature matrix of all students and their economic hardship scores as input, and the economic hardship level as pseudo-label. Use the trained CART decision tree to predict economic hardship for all students in the current group; By summarizing all the predicted economically disadvantaged students from various groups, we can obtain the output of all suspected economically disadvantaged students in colleges and universities.
[0014] Secondly, a college student financial difficulty identification system based on the EW-KMeans algorithm is provided, including: The data collection and aggregation module is used to collect all students' consumption data and library access control data, and to statistically analyze all characteristic indicators of each student. The weight determination module is used to calculate the weight of each feature indicator based on all the feature indicators of each student using the entropy weight method, and to correct the calculated weight of each indicator by introducing an entropy weight truncation and redistribution method. The classification module is used to construct a weighted feature matrix for each student based on the corrected weights of each indicator and the statistical feature indicators of each student. After determining the optimal number of clusters K using the elbow method, it clusters all students based on the weighted feature matrices of all students to obtain several student groups. The segmentation and identification module is used to calculate the economic hardship score of each student in each group based on the relative benchmark of their characteristic indicators and classify them into levels. Then, a decision tree is used to perform rule mining within each group to identify all suspected economically disadvantaged students in each group.
[0015] Thirdly, a computer device is provided, including a processor and a memory; wherein, when the processor executes a computer program stored in the memory, it implements the steps of the method for identifying economically disadvantaged college students based on the EW-KMeans algorithm as described in any one of the first aspects.
[0016] Fourthly, a computer-readable storage medium is provided for storing a computer program; when the computer program is executed by a processor, it implements the steps of the method for identifying economically disadvantaged college students based on the EW-KMeans algorithm as described in any one of the first aspects.
[0017] Compared with the prior art, the beneficial effects of the present invention are: (1) First, in the feature weighting stage before clustering, this invention introduces an entropy weight truncation and redistribution mechanism. By setting an upper limit for weights, the weights exceeding the upper limit are evenly distributed to features below the threshold. This avoids a single or a few features from dominating the clustering distance due to excessive variability, and prevents the clustering results from over-reliance on noise or redundant features. At the same time, it preserves the distinguishing ability of each feature, making the clustered groups have a more balanced and reliable behavioral description, providing a high-quality group segmentation basis for subsequent economic hardship identification. Second, this invention divides students into subgroups with similar behavioral patterns through clustering, and then constructs a relative comparison benchmark based on the consumption level of each group to calculate the relative economic hardship of individuals. This strategy allows the assessment of economic hardship to break free from the constraints of absolute consumption values, achieving "comparison within the same group," significantly reducing misjudgments and omissions caused by differences in group styles, and improving the fairness of the identification results. Finally, this invention uses a two-level architecture of coarse clustering and detailed decision tree subdivision to mine economic hardship behavior rules for different groups. In particular, for hidden economic hardship groups that are difficult to identify by traditional models, such as "free agents" (low campus consumption and low participation) and "academic high achievers" (simple consumption but high learning investment), the decision tree can learn targeted feature combinations within each group, thereby significantly improving the accuracy of identifying hidden economic hardship.
[0018] (2) The CART decision tree of this invention uses the Gini coefficient as the splitting criterion and produces a set of intuitive if-then classification rules. These rules are easy for auditors to understand and verify, avoiding the trust crisis caused by "black box" models. At the same time, the rules can be updated periodically based on new data, realizing dynamic adjustment and precise assistance for the identification of economically disadvantaged students. Attached Figure Description
[0019] Figure 1This is a flowchart of the steps of the method for identifying economically disadvantaged college students based on the EW-KMeans algorithm of the present invention; Figure 2 This is a comparison chart of the clustering results using the present invention and existing clustering methods. Figure 2 (a) shows the PCA results of clustering after calculating the weighted feature matrix using the entropy weight truncation and redistribution method introduced in this invention. Figure 2 (b) in the figure shows the PCA results of clustering after calculating the weighted feature matrix using existing technology with equal weights. Detailed Implementation
[0020] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be used to limit the scope of protection of the present invention. It should be noted that the term "comprising" and any variations thereof in the specification, claims and the above-mentioned drawings of the present invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or devices.
[0021] Example 1: like Figure 1 As shown, a method for identifying economically disadvantaged college students based on the EW-KMeans algorithm is provided, including: Step S1: Collect all students' consumption data and library access control data, and compile all characteristic indicators for each student, including consumption indicators and library indicators.
[0022] Export student data for the statistical period from the campus card management system, including fields such as student ID, spending amount, and spending time; export access control data for students entering the library from the access control management system, including fields such as student ID, access time, and identification method; store the exported data in a readable and writable format such as Excel or CSV, with each event recorded as a record.
[0023] After preprocessing the data, all characteristic indicators for each student are calculated.
[0024] Consumption indicators include: number of transaction days, average daily consumption, average transaction amount, and breakfast rate; library indicators include: the presence and intensity of library activities. The existence characteristics and intensity of library behavior are determined based on the number of times students access the library during the statistical period.
[0025] (1) Number of transaction days: The number of valid days with consumption records within the statistical period: ; in, Indicates the first The number of trading days for each student within the statistical period. Indicates the first The first student Previous purchase record Indicates the time and date of extraction. This represents the statistics after removing duplicates.
[0026] (2) Average daily consumption: The ratio of total transaction amount to the number of transaction days within the statistical period. This indicator reflects the students' daily consumption level. ; in, Indicates the first Average daily spending per student Indicates the first The first student t Daily spending amount.
[0027] (3) Amount of a single transaction: The ratio of the total transaction amount to the total number of transactions reflects the student's consumption level in a single transaction.
[0028] (4) Breakfast rate: An important indicator reflecting students' daily routines and lifestyles, used to measure the frequency of breakfast consumption within the effective transaction days. The default time for breakfast consumption is between 5:00 AM and 10:00 AM. ; in, Indicates the first i The number of valid days for each student's breakfast consumption is used to determine whether it falls within the breakfast time period. If it falls within the breakfast time period, the value is assigned as 1; otherwise, it is assigned as 0. Finally, the total number of days that students have transaction records during the breakfast time period is counted.
[0029] ; in, Indicates the first i Breakfast rate per student.
[0030] (5) Number of times the library access card is accessed: ; in, Indicates the first i The number of times each student accesses the library during a given period. Indicates the first The first student Library access control records.
[0031] Library access control data exhibits significant variability, with a large number of access records showing zero entries, forming a sharp peak on the left side of the distribution. Simultaneously, a small number of access records far exceed the average, forming a long tail, which can negatively impact the final evaluation. This paper proposes a divide-and-conquer strategy for access control data, employing a dual-channel decoupled preprocessing method. This method uses two dimensions to represent student learning behavior, decoupling library access counts into the presence and intensity of library activities.
[0032] The existential characteristics of library behavior are: ; in, For the first Existential characteristics of students' library behavior. , For the number of students, For the first The number of times a student accesses the library during the statistical period; The intensity of library behavior is: ; in, , For all The average value, For all The standard deviation.
[0033] Table 1 provides an example of statistical analysis of all characteristic indicators of a student.
[0034] Table 1: Examples of feature indicators for multiple samples.
[0035]
[0036] Step S2: Based on all the characteristic indicators of each student, the weight of each characteristic indicator is calculated using the entropy weight method, and the calculated weight of each indicator is corrected by introducing the entropy weight truncation and redistribution method.
[0037] Step S2 specifically includes the following steps (steps D1-D3): Step D1: Calculate the initial weight of each feature indicator using the entropy weight method, and set the upper limit of the weight threshold based on the reciprocal of the number of feature indicators. .
[0038] ;in, This represents the number of feature indicators. For the specific steps of calculating the initial weight of each feature indicator using the entropy weight method, please refer to existing techniques.
[0039] Step D2: Determine whether the current weights of all feature indicators are less than the upper limit of the weight threshold. If yes, output the current weights of all feature metrics; otherwise, start from a value greater than the upper limit of the weight threshold. From all the current weights, select the current weight of any feature indicator, and adjust the current weights of some feature indicators according to the following formula.
[0040] ; in, From greater than the upper limit of the weight threshold The first selected from The current weights of each feature indicator, For the revised first The current weights of each feature indicator, Less than the upper limit of the weight threshold The The current weights of each feature indicator, For the revised first The current weights of each feature indicator, This represents the sum of the excess amounts of all features that exceed the threshold. Less than the upper limit of the weight threshold The current number of weights; Step D3: Repeat step D2 until the current weights of all feature indicators are less than the upper limit of the weight threshold. .
[0041] The entropy weight truncation and redistribution method in step S2 retains the sensitivity of the entropy weight method to data differences while avoiding evaluation imbalance caused by excessive fluctuations in individual indicators.
[0042] Step S3: Based on the corrected weights of each indicator and the statistical characteristic indicators of each student, construct a weighted feature matrix for each student, and use the elbow method to determine the optimal number of clusters K. Then, cluster all students based on the weighted feature matrices of all students to obtain several student groups.
[0043] The weighted feature matrix for each student is represented as follows: ; in, For the first The weighted feature matrix of each student For the first The first student Each weighted feature indicator For the first The weights of each feature indicator after correction For the first The first student Each characteristic indicator.
[0044] The specific steps for determining the optimal number of clusters K using the elbow method are based on existing techniques. All students are clustered based on the weighted feature matrix of all students to obtain several student groups. Existing clustering algorithms, such as the K-Means algorithm, can be used.
[0045] pass Figure 2 It can be seen that by using the entropy weight truncation and redistribution method of the present invention to correct the weights and then perform clustering, the clustering quality is better and the group division is more effective.
[0046] A specific example of a group label is given in Table 2: Table 2: Examples of group labels.
[0047]
[0048] Step S4: For each student group, calculate the economic hardship score of each student in the group based on the relative benchmark of their consumption indicators and classify them into levels. Then, use decision trees to perform rule mining within each group to identify all suspected economically disadvantaged students in each group.
[0049] Step S4 is as follows: S41: For each student group, calculate the 75th percentile of all their consumption indicators, and use the 75th percentile as the benchmark value for the current student group in terms of current consumption indicators.
[0050] S42: Calculate each student's individual economic hardship score according to the following formula; ; Among them, among them, For the first The student in the first The one-way economic difficulty score of each consumption indicator, For the first The first student A weighted consumption indicator (consumption indicator among weighted characteristic indicators). For the first The first among the student groups The benchmark value of the first consumption indicator, i.e., the first The first among the student groups The 75th percentile of each characteristic indicator. Choosing the 75th percentile avoids the influence of extremes; if a student's spending is particularly high, it will not raise the comparison benchmark.
[0051] S43: By weighting and aggregating all individual economic hardship scores of students, a student's economic hardship score is obtained, and each student is classified into a level of economic hardship based on their economic hardship score.
[0052] ; in, For the first Each student's score on their level of financial hardship For the quantity of consumption indicators, for The preset fusion weights.
[0053] Set a threshold for the level of economic hardship, for example: For particularly difficult For those with general economic difficulties, As a marginalized group, It is not economic hardship.
[0054] S44: For each group, a CART decision tree is trained using the weighted feature matrix of all students and the students' economic hardship scores as inputs, and the economic hardship level as pseudo-labels.
[0055] S45: Use the trained CART decision tree to predict economic hardship for all students in the current group.
[0056] The structure of CART decision trees can refer to existing technologies. CART decision trees are based on the Gini coefficient and can be used to predict economic hardship, enabling secondary subdivision within a population. Each population corresponds to one CART decision tree.
[0057] S46: Summarize all the predicted economically disadvantaged students from each group to obtain the output of all suspected economically disadvantaged students in colleges and universities, for reviewers to verify and further evaluate.
[0058] For the groups categorized in Table 2, the results of statistical analysis of economically disadvantaged students in each group using step S4 of this invention are presented.
[0059] Table 3: Statistical results of suspected economically disadvantaged students in various groups.
[0060]
[0061] Statistical results show that there are significant differences in the incidence of economic hardship among different clusters: Group 4 (freelancers) has the highest incidence of economic hardship (11.1%), followed by Group 3 (academic high achievers) (9.2%), while Group 1 (low-desire type) has only 4.4%, and is not the group with the highest proportion of economic hardship.
[0062] Traditional methods of identifying economic hardship solely based on low consumption are flawed. Low consumption does not directly equate to economic hardship; some students with low desires are simply frugal in their lifestyle. The marginalized group is a high-risk area for hidden economic hardship; their low campus spending and low library participation are easily missed by traditional methods. This model achieves accurate identification through cluster modeling. A large number of motivated students with economic hardship exist within the high-achieving student group; they dedicate their main energy to studying and have simple consumption patterns, making them an important part of the economically disadvantaged student population.
[0063] Example 2: A college student financial aid identification system based on the EW-KMeans algorithm includes: The data collection and aggregation module is used to collect all students' consumption data and library access control data, and to statistically analyze all characteristic indicators of each student. The weight determination module is used to calculate the weight of each feature indicator based on all the feature indicators of each student using the entropy weight method, and to correct the calculated weight of each indicator by introducing an entropy weight truncation and redistribution method. The classification module is used to construct a weighted feature matrix for each student based on the corrected weights of each indicator and the statistical feature indicators of each student. After determining the optimal number of clusters K using the elbow method, it clusters all students based on the weighted feature matrices of all students to obtain several student groups. The segmentation and identification module is used to calculate the economic hardship score of each student in each group based on the relative benchmark of their characteristic indicators and classify them into levels. Then, a decision tree is used to perform rule mining within each group to identify all suspected economically disadvantaged students in each group.
[0064] Example 3: The present invention provides a computer device, including a processor and a memory; wherein, when the processor executes the computer program stored in the memory, it implements the steps of the above-mentioned method for identifying economically disadvantaged college students based on the EW-KMeans algorithm.
[0065] For a more detailed explanation of the above method, please refer to the relevant content disclosed in the foregoing embodiments, which will not be repeated here.
[0066] Example 4: This invention provides a computer-readable storage medium for storing a computer program; when the computer program is executed by a processor, it implements the steps of the above-described method for identifying economically disadvantaged college students based on the EW-KMeans algorithm.
[0067] For a more detailed explanation of the above method, please refer to the relevant content disclosed in the foregoing embodiments, which will not be repeated here.
[0068] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. The systems, devices, and storage media disclosed in the embodiments are described simply because they correspond to the methods disclosed in the embodiments; relevant details can be found in the method section.
[0069] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.
[0070] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.
Claims
1. A method for identifying economically disadvantaged college students based on the EW-KMeans algorithm, characterized in that, include: Collect all students' consumption data and library access control data, and compile all characteristic indicators for each student, including consumption indicators and library indicators; Based on all the characteristic indicators of each student, the entropy weight method is used to calculate the weight of each characteristic indicator, and the entropy weight truncation and redistribution method is introduced to correct the calculated weight of each indicator. Based on the corrected weights of each indicator and the statistical characteristic indicators of each student, a weighted feature matrix is constructed for each student. After determining the optimal number of clusters K using the elbow method, all students are clustered based on the weighted feature matrices of all students to obtain several student groups. For each student group, the economic hardship score of each student in the group is calculated and classified based on the relative benchmark of their consumption indicators. Then, decision trees are used to mine rules within each group to identify all suspected economically disadvantaged students in each group.
2. The method for identifying economically disadvantaged college students based on the EW-KMeans algorithm according to claim 1, characterized in that, The consumption indicators include: number of transaction days, average daily consumption, amount per transaction, and breakfast rate; the library indicators include: library behavior presence characteristics and library behavior intensity; the library behavior presence characteristics and library behavior intensity are determined based on the number of times a student accesses the library during the statistical period.
3. The method for identifying economically disadvantaged college students based on the EW-KMeans algorithm according to claim 2, characterized in that, The existence characteristics of the library behavior are as follows: ； in, For the first Existential characteristics of students' library behavior. , For the number of students, For the first The number of times a student accesses the library during the statistical period; The intensity of the library behavior is: ； in, , For all The average value, For all The standard deviation.
4. The method for identifying economically disadvantaged college students based on the EW-KMeans algorithm according to claim 1, characterized in that, Based on all the characteristic indicators of each student, the entropy weight method is used to calculate the weight of each characteristic indicator, and the calculated indicator weights are corrected by introducing an entropy weight truncation and redistribution method, specifically: Step D1: Calculate the initial weight of each feature indicator using the entropy weight method, and set the upper limit of the weight threshold based on the reciprocal of the number of feature indicators. ; Step D2: Determine whether the current weights of all feature indicators are less than the upper limit of the weight threshold. If yes, output the current weights of all feature metrics; otherwise, start from a value greater than the upper limit of the weight threshold. From all the current weights, select the current weight of any feature indicator, and adjust the current weights of some feature indicators according to the following formula; ； in, From greater than the upper limit of the weight threshold The first selected from The current weights of each feature indicator, For the revised first The current weights of each feature indicator, Less than the upper limit of the weight threshold The The current weights of each feature indicator, For the revised first The current weights of each feature indicator, This represents the sum of the excess amounts of all features that exceed the threshold. Less than the upper limit of the weight threshold The current number of weights; Step D3: Repeat step D2 until the current weights of all feature indicators are less than the upper limit of the weight threshold. .
5. The method for identifying economically disadvantaged college students based on the EW-KMeans algorithm according to claim 1, characterized in that, The weighted feature matrix for each student is represented as follows: ； in, For the first The weighted feature matrix of each student For the first The first student Each weighted feature indicator For the first The weights of each feature indicator after adjustment For the first The first student Each characteristic indicator.
6. The method for identifying economically disadvantaged college students based on the EW-KMeans algorithm according to claim 1, characterized in that, The calculation of economic hardship scores for each student within the group, based on a relative benchmark of their consumption indicators, and the subsequent classification into levels, is as follows: For each student group, calculate the 75th percentile of all their consumption indicators, and use the 75th percentile as the benchmark value for the current student group in terms of current consumption indicators. Calculate each student's individual economic hardship score according to the following formula; ； in, For the first The student in the first The one-way economic difficulty score of each consumption indicator, For the first The first student A weighted consumption indicator, For the first The first among the student groups The benchmark value for each consumption indicator; By weighting and aggregating all individual economic hardship scores, a student's economic hardship score is obtained: ； in, For the first Each student's score on their level of financial hardship For the quantity of consumption indicators, for The fusion weight; Each student is classified into different levels of financial hardship based on their financial hardship score.
7. The method for identifying economically disadvantaged college students based on the EW-KMeans algorithm according to claim 6, characterized in that, The process of using decision trees to perform rule mining within each group and identifying all suspected economically disadvantaged students in each group specifically involves: For each group, a CART decision tree is trained using the weighted feature matrix of all students and their economic hardship scores as input, and the economic hardship level as pseudo-label. Use the trained CART decision tree to predict economic hardship for all students in the current group; By summarizing all the predicted economically disadvantaged students from various groups, we can obtain the output of all suspected economically disadvantaged students in colleges and universities.
8. A college student financial aid identification system based on the EW-KMeans algorithm, characterized in that, include: The data collection and aggregation module is used to collect all students' consumption data and library access control data, and to statistically analyze all characteristic indicators of each student. The weight determination module is used to calculate the weight of each feature indicator based on all the feature indicators of each student using the entropy weight method, and to correct the calculated weight of each indicator by introducing an entropy weight truncation and redistribution method. The classification module is used to construct a weighted feature matrix for each student based on the corrected weights of each indicator and the statistical feature indicators of each student. After determining the optimal number of clusters K using the elbow method, it clusters all students based on the weighted feature matrices of all students to obtain several student groups. The segmentation and identification module is used to calculate the economic hardship score of each student in each group based on the relative benchmark of their characteristic indicators and classify them into levels. Then, a decision tree is used to perform rule mining within each group to identify all suspected economically disadvantaged students in each group.
9. A computer device, characterized in that, It includes a processor and a memory; wherein, when the processor executes the computer program stored in the memory, it implements the steps of the method for identifying economically disadvantaged college students based on the EW-KMeans algorithm as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, Used to store computer programs; when the computer programs are executed by a processor, they implement the steps of the method for identifying economically disadvantaged college students based on the EW-KMeans algorithm as described in any one of claims 1-7.