Data analysis-based elderly study decision-making system

Abstract

The invention relates to the technical field of data analysis, in particular to an old people research decision system based on data analysis, which comprises a learning behavior analysis module, a learning effect monitoring module, a research content adaptation module and a decision optimization support module. According to the method, the learning activities of the elderly users are monitored and analyzed in real time through cloud computing and edge computing, accurate collection of data is improved, instant elimination of abnormal data is enhanced, the quality of data analysis is ensured, diversity of user requirements is effectively coped with through automatic generation of targeted learning content, and the user experience is improved. According to the method, the education content is more attractive, the learning experience of the user is enhanced, the burden of the user during learning content selection is reduced through the optimized recommendation process, the decision-making efficiency and satisfaction degree are improved, the education plan can be adjusted more accurately, old people are helped to make more suitable learning choices, and therefore the overall education effect is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a research and study decision-making system for the elderly based on data analysis. Background Art

[0002] The field of data analysis technology covers the entire process from basic data collection and processing to complex analysis and prediction, including the application of data mining, machine learning, statistical analysis and big data technology, aiming to extract valuable information from massive data to support decision-making. Data analysis plays a core role in many fields such as business strategy, health science, financial market forecasting and customer behavior analysis. With the development of technology, data analysis is increasingly relying on automated tools and algorithms to improve processing efficiency and analysis accuracy.

[0003] Among them, the elderly study and research decision-making system based on data analysis refers to the use of data analysis technology to support and optimize the decision-making process of elderly education and learning activities. The system integrates user behavior data, learning outcome data, and environmental factor data, and provides customized learning suggestions and activity arrangements by analyzing data. Its uses include helping education providers analyze the learning needs and preferences of the elderly, thereby designing more appropriate educational courses and activities, and also helping the elderly themselves make more appropriate learning choices.

[0004] Traditional technology has difficulty capturing subtle differences in user behavior, resulting in difficulty in analyzing and meeting the learning needs of the elderly in the provision of educational content. Data processing delays and lack of real-time feedback are also a major shortcoming, which is not conducive to providing timely updated learning content and cannot meet the immediate learning needs of elderly users. For example, untimely data updates result in recommended content no longer being suitable for the user's actual learning status, affecting learning motivation and effectiveness. Existing technologies rely on relatively static algorithms to generate targeted learning content and lack dynamic adaptability, which leads to a misalignment between users and the best learning materials in a rapidly changing educational environment, thereby reducing learning satisfaction and efficiency. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose a research and study decision-making system for the elderly based on data analysis.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: a research and study decision-making system for the elderly based on data analysis, the system comprising:

[0007] The learning behavior analysis module is based on the learning activity records of elderly users. Through cloud computing and edge computing, it collects learning time, content browsing and interaction frequency, creates corresponding timestamps for each activity, and records the attributes and time information of each learning activity to generate an overview of behavioral characteristics.

[0008] The learning effect monitoring module uses data cleaning technology to remove abnormal records based on the behavioral feature overview, conducts linear analysis on the learning progress, quantitatively evaluates the user's understanding depth and memory retention rate, and generates quantitative analysis information on the effect;

[0009] The study content adaptation module identifies matching learning topics and materials based on the quantitative analysis information of the effects, combined with user preferences and historical behaviors, automatically generates targeted study content through the NLP model, and formats the content into text, images or videos to obtain content matching indicators;

[0010] The decision optimization support module analyzes the user's learning effects and behavior patterns based on the content matching indicators, optimizes the research and study content recommendation process, reduces the user's selection range, and matches the most suitable content options to obtain detailed content recommendation results.

[0011] The present invention is improved in that the steps of obtaining the behavioral feature overview are specifically as follows:

[0012] Based on the learning activity records of elderly users, the learning data of elderly users are collected through cloud computing and edge computing, including the length of learning time, content browsing volume and interaction frequency, using the formula:

[0013] D i =(T i ,V i ,F i ,t i )

[0014] Get the data tuple D of the i-th activity i , where T i represents the learning time length of the ith activity, V i represents the number of content views of the i-th activity, F i represents the interaction frequency of the ith activity, t i Represents the timestamp of the i-th activity;

[0015] Aggregate the data tuples and calculate the weighted average of the activities using the formula:

[0016]

[0017] and

[0018]

[0019] and

[0020]

[0021] Get the weighted average length of study time Content views and interaction frequency Among them, w T 、w V and w F Represent the weight parameters of learning time length, content browsing volume and interaction frequency, T i represents the learning time length of the ith activity, V i represents the number of content views of the i-th activity, F i represents the interaction frequency of the ith activity;

[0022] Based on the weighted average learning time length, content browsing volume and interaction frequency, combined with the activity time attribute, an overview of the behavioral characteristics is obtained.

[0023] The present invention is improved in that the step of performing linear analysis on the learning progress is specifically:

[0024] Based on the behavioral characteristics overview, the cleaned learning data is integrated to extract each user's learning time and number of completed courses using the formula:

[0025]

[0026] Get the preliminary learning progress data PD, where: is the learning time of the ith data point, is the corresponding number of completed courses;

[0027] Based on the preliminary learning progress data, a linear regression analysis is performed to calculate the best fitting straight line parameters using the formula:

[0028]

[0029] and

[0030]

[0031] Get the linear model parameters, where a d is the slope, b d is the intercept, n d is the total number of data points, is the learning time of the ith data point, is the corresponding number of completed courses;

[0032] Using the linear model parameters, the learning progress of future or unfinished data is predicted using the formula:

[0033] c predicted =a d ·t d +b d

[0034] Get the predicted learning progress result, where a d is the slope, b d is the intercept, c predicted It is based on the user's future learning time t d The predicted number of course completions.

[0035] The present invention is improved in that the step of obtaining the effect quantitative analysis information is specifically as follows:

[0036] Based on the learning progress information, the predicted learning outcomes of each user are calculated, and the estimated values ​​of the understanding depth and memory retention rate of each user are extracted to obtain a preliminary evaluation result set;

[0037] Based on the preliminary evaluation result set, the prediction results of each user are analyzed using the formula:

[0038]

[0039] Calculate the depth of understanding and memory retention rate to obtain the quantitative evaluation result QR and obtain the quantitative analysis information of the effect, where R s is the user's predicted learning outcome, α r is the sensitivity coefficient, γ r is the linear influence coefficient, δ r is the smoothing parameter.

[0040] The present invention is improved in that the steps of identifying the learning topics and materials are specifically as follows:

[0041] Based on the quantitative analysis information of the effect, the learning outcomes of each user are extracted, and the adaptability score of each learning topic is calculated by combining the user preference data and historical behavior, using the formula:

[0042]

[0043] Get the user's topic matching score ES i , where EG i Represents the learning outcome score, EP i Indicates preference score, EH i represents the historical behavior score, a x , b x and c x is the weight parameter;

[0044] The topic matching scores of the users are screened, and the learning topics with the best scores are selected. The NLP model is used to automatically extract the learning materials associated with the topics from the content library, using the formula:

[0045]

[0046] Get the matching learning material list LX, where NLP_score j is the relevance score given by the NLP model, is the weight coefficient of the material relevance score, n x is the material quantity;

[0047] Based on the matching learning material list, the material selection is optimized to match the educational needs, using the formula:

[0048]

[0049] Get the list of learning topics and materials, where OX represents optimization score, Diversity_score k and Depth_score k are the diversity and depth scores of the materials, λ x and μ x is the influence that regulates diversity and depth, and m is the number of materials that have been screened and evaluated.

[0050] The present invention is improved in that the step of obtaining the content matching index is specifically as follows:

[0051] Collect and organize user interaction data and user feedback data, initialize the calculation of matching indicators, and use the formula:

[0052]

[0053] Get the initial indicator set IY, where UY is user interaction data and FY is user feedback data. and is a weight parameter used to control the impact of the data source on the score;

[0054] Based on the preliminary indicator set, the matching indicators are iteratively adjusted and refined by comparing with the quality standards, using the formula:

[0055] MY=ρ Y ·(IY+QY)

[0056] Get the content matching index MY, where QY is the quality control parameter, IY is the preliminary index set, ρ Y It is the adjustment coefficient used to balance the relationship between the initial indicator and the quality standard.

[0057] The present invention is improved in that the steps of analyzing the user's learning effect and behavior pattern are specifically as follows:

[0058] Based on the content matching index, the user interaction data is integrated to calculate the interaction score using the formula:

[0059] Score interact=Fw t ·T L +Fw c ·C L

[0060] Get the user interaction score Score interact , where T L is the residence time, C L is the number of clicks, Fw t and Fw c They are the weight coefficients of time and click respectively;

[0061] Based on the user interaction scores, learning behavior patterns are identified using the formula:

[0062]

[0063] The relationship between the interaction score and the learning outcome is expressed as Slope, where x i is the individual value of the interaction score, y i is the corresponding learning outcome, and is the average of the interaction score and the learning outcome.

[0064] The present invention is improved in that the steps of obtaining the refined content recommendation results are specifically as follows:

[0065] According to user interaction data, adjust the recommendation logic of research content, optimize the pertinence and accuracy, and analyze the relationship between user behavior and content preferences. Use the formula:

[0066]

[0067] Get the optimized recommendation logic, where GE x represents the user feature vector, θ is the parameter vector of the logistic regression model, P(y x =1|GE x ) is the given user feature GE x Next, the predicted probability of liking a certain content, θ T GE x is the dot product of the parameter vector and the eigenvector;

[0068] Based on the optimized recommendation logic, the recommendation score is calculated and the content ranking is determined using the formula:

[0069] DScore = α B PDY+β B ·RLC

[0070] The refined content recommendation result DScore is obtained, where PDY represents the popularity of the content, RLC is the correlation between the content and the user's preference, and α B and β B is a tuning parameter.

[0071] Compared with the prior art, the advantages and positive effects of the present invention are:

[0072] In the present invention, real-time monitoring and analysis of the learning activities of elderly users through cloud computing and edge computing not only improves the accuracy of data collection, but also strengthens the immediate elimination of abnormal data, ensuring the quality of data analysis. The automatic generation of targeted learning content effectively responds to the diversity of user needs, making educational content more attractive and enhancing the user's learning experience. The optimized recommendation process reduces the burden on users in selecting learning content, improves decision-making efficiency and satisfaction, and enables more accurate adjustment of education plans to help the elderly make learning choices that are more suitable for them, thereby improving the overall education effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 is a system flow chart of the present invention;

[0074] Figure 2 A flowchart for obtaining an overview of behavioral features in the present invention;

[0075] Figure 3 A flowchart of the linear analysis of learning progress in the present invention;

[0076] Figure 4 This is a flow chart for obtaining the effect quantitative analysis information in the present invention;

[0077] Figure 5 A flow chart for identifying learning topics and materials in the present invention;

[0078] Figure 6 This is a flow chart for obtaining content matching indicators in the present invention;

[0079] Figure 7 It is a flow chart for analyzing the learning effect and behavior pattern of users in the present invention;

[0080] Figure 8 This is a flowchart for obtaining the refined content recommendation results in the present invention. DETAILED DESCRIPTION

[0081] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0082] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating positions or positional relationships, are based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0083] Example

[0084] See also Figure 1 The present invention provides a technical solution: a research and study decision-making system for the elderly based on data analysis includes:

[0085] The learning behavior analysis module is based on the learning activity records of elderly users. Through cloud computing and edge computing, it collects learning time, content browsing and interaction frequency, creates corresponding timestamps for each activity, and records the attributes and time information of each learning activity to generate an overview of behavioral characteristics.

[0086] The learning effect monitoring module uses data cleaning technology to remove abnormal records based on the behavioral feature overview, conducts linear analysis of learning progress, quantitatively evaluates user understanding depth and memory retention rate, and generates quantitative analysis information of the effect;

[0087] The research and study content adaptation module identifies matching learning topics and materials based on quantitative effect analysis information, combined with user preferences and historical behaviors, automatically generates targeted research and study content through the NLP model, and formats the content into text, images or videos to obtain content matching indicators;

[0088] The decision optimization support module analyzes user learning effects and behavior patterns based on content matching indicators, optimizes the research and study content recommendation process, reduces the user's selection range, and matches the most suitable content options to obtain detailed content recommendation results.

[0089] The behavioral characteristics overview includes activity period, interaction type, and visit frequency. The quantitative analysis information of the effect includes evaluation criteria, improvement points, and monitoring indicators. The content matching indicators include content coverage, material matching, and content diversity. The detailed content recommendation results include recommendation adaptation rate and user feedback information.

[0090] See also Figure 2 , the steps to obtain the behavioral feature overview are as follows:

[0091] Based on the learning activity records of elderly users, the learning data of elderly users are collected through cloud computing and edge computing, including the length of learning time, content browsing volume and interaction frequency, using the formula:

[0092] D i =(T i ,V i ,F i ,t i )

[0093] Get the data tuple D of the i-th activity i , where T i represents the learning time length of the ith activity, V i represents the number of content views of the i-th activity, F i represents the interaction frequency of the ith activity, t i Represents the timestamp of the i-th activity;

[0094] Aggregate the data tuples and calculate the weighted average of the activities using the formula:

[0095]

[0096] and

[0097]

[0098] and

[0099]

[0100] Get the weighted average length of study time Content views and interaction frequency Among them, w T 、w V and w F Represent the weight parameters of learning time length, content browsing volume and interaction frequency, T i represents the learning time length of the ith activity, V i represents the number of content views of the i-th activity, F i represents the interaction frequency of the ith activity;

[0101] Based on the weighted average learning time length, content views and interaction frequency, combined with the activity time attributes, an overview of behavioral characteristics is obtained.

[0102] Assume that during a learning activity, a user:

[0103] T i =30 minutes;

[0104] Vi =15 pages;

[0105] F i = 10 interactions;

[0106] t i It is 10:30 on August 1, 2024.

[0107] D i =(30, 15, 10, "2024-08-0110:30"), this data tuple represents the specific data of a learning activity, including learning time, page views, interaction frequency and activity timestamp.

[0108] Suppose there are three activity data:

[0109] first:

[0110] T1=30, V1=15, F1=10;

[0111] The weights are:

[0112]

[0113] Second time:

[0114] T2=40, V2=20, F2=15;

[0115] The weights are:

[0116]

[0117] Third time:

[0118] T3=50, V3=25, F3=5;

[0119] The weights are:

[0120]

[0121] calculate

[0122]

[0123] The results showed a weighted average learning time of 39 minutes, an average of 21.5 pages viewed, and an average of 9.5 interactions.

[0124] See also Figure 3 , the steps for linear analysis of learning progress are as follows:

[0125] Based on the behavioral feature overview, the cleaned learning data is integrated to extract each user's learning time and number of completed courses using the formula:

[0126]

[0127] Get the preliminary learning progress data PD, where: is the learning time of the ith data point, is the corresponding number of completed courses;

[0128] Based on the preliminary learning progress data, linear regression analysis was performed to calculate the best fitting straight line parameters using the formula:

[0129]

[0130] and

[0131]

[0132] Get the linear model parameters, where a d is the slope, b d is the intercept, n d is the total number of data points, is the learning time of the ith data point, is the corresponding number of completed courses;

[0133] Using the linear model parameters, predict the learning progress of future or unfinished data using the formula:

[0134] c predicted =a d ·t d +b d

[0135] Get the predicted learning progress result, where a d is the slope, b d is the intercept, c predicted It is based on the user's future learning time t d The predicted number of course completions.

[0136] Suppose there are three data points, representing the study time and the number of courses completed:

[0137]

[0138] Compute the sum of the terms:

[0139]

[0140] n d =3

[0141] Slope a d Calculation:

[0142]

[0143] Intercept b d Calculation:

[0144]

[0145] If we want to predict the learning time t d =Number of courses completed for 4 hours:

[0146] c predicted =1.5·4+0.67=6+0.67=6.67

[0147] If a student studies for 4 units, the predicted number of courses completed is approximately 6.67.

[0148] See also Figure 4 , the specific steps for obtaining the effect quantitative analysis information are as follows:

[0149] Based on the learning progress information, the predicted learning outcomes of each user are calculated, from which the estimated values ​​of each user's understanding depth and memory retention rate are extracted to obtain a preliminary evaluation result set;

[0150] Based on the preliminary evaluation result set, the prediction results for each user are analyzed using the formula:

[0151]

[0152] Calculate the depth of understanding and memory retention rate to obtain the quantitative evaluation result QR and obtain the quantitative analysis information of the effect, where R s is the user's predicted learning outcome, α r is the sensitivity coefficient, γ r is the linear influence coefficient, δ r is the smoothing parameter.

[0153] Calculation process:

[0154] For Users

[0155] Compute the logarithmic term:

[0156]

[0157] Calculate the total score:

[0158] QR1=0.5·2.197+0.3·0.8≈1.098+0.24=1.338

[0159] For Users

[0160] Compute the logarithmic term:

[0161]

[0162] Calculate the total score:

[0163] QR2=0.5·2.773+0.3·1.5≈1.387+0.45=1.837

[0164] For Users

[0165] Compute the logarithmic term:

[0166]

[0167] Calculate the total score:

[0168] QR3=0.5·1.099+0.3·0.2≈0.549+0.06=0.609

[0169] QR1=1.338 indicates that user 1 has a higher depth of understanding and memory retention rate;

[0170] QR2=1.837 indicates that User 2’s depth of understanding and memory retention rate are very high;

[0171] QR3=0.609 indicates that user 3 has a lower depth of understanding and memory retention rate.

[0172] See also Figure 5 , the steps for identifying learning topics and materials are as follows:

[0173] Based on the quantitative analysis information of the effect, the learning outcomes of each user are extracted, and the adaptability score of each learning topic is calculated by combining the user preference data and historical behavior. The formula is:

[0174]

[0175] Get the user's topic matching score ES i , where EG i Represents the learning outcome score, EP i Indicates preference score, EH i represents the historical behavior score, a x , b x and c x is the weight parameter;

[0176] Filter the user's topic matching score, select the learning topic with the highest score, and use the NLP model to automatically extract the learning materials associated with the topic from the content library using the formula:

[0177]

[0178] Get the matching learning material list LX, where NLP_scorej is the relevance score given by the NLP model, is the weight coefficient of the material relevance score, n x is the material quantity;

[0179] Based on the matching learning material list, optimize the material selection to match the educational needs, using the formula:

[0180]

[0181] Get the list of learning topics and materials, where OX represents optimization score, Diversity_score k and Depth_score k are the diversity and depth scores of the materials, λ x and μ x is the influence that regulates diversity and depth, and m is the number of materials that have been screened and evaluated.

[0182] Assumptions:

[0183] a x =0.5, b x =0.3, c x =0.2;

[0184] For Topic I:

[0185] EG i =85,EP i =90, EH i =75;

[0186] Calculation process:

[0187]

[0188] The overall fit score for the subject is 84.5, reflecting the subject's suitability based on learning outcomes, preferences, and historical behavior.

[0189] Assume φ x =1.5, assuming there are three NLP scores for the material:

[0190] 95, 88, 82.

[0191] Calculation process:

[0192] LX=1.5·(95+88+82)

[0193] LX=1.5·265=397.5

[0194] It means that after weighting, the total score of the screened materials is 397.5, which quantifies the matching degree between the materials and the learning topics.

[0195] Assumptions:

[0196] λ x =2,μ x =1;

[0197] Suppose there are two materials with diversity and depth scores:

[0198] (80, 70) and (90, 85).

[0199] Calculation process:

[0200] OX=(2·80+1·70)+(2·90+1·85)

[0201] OX = 160 + 70 + 180 + 85 = 495

[0202] It means that after weight adjustment, the total optimization score of the material is 495, which is used to ultimately decide the allocation and optimization strategy of educational resources.

[0203] See also Figure 6 , the steps to obtain the content matching index are as follows:

[0204] Collect and organize user interaction data and user feedback data, initialize the calculation of matching indicators, and use the formula:

[0205]

[0206] Get the initial indicator set IY, where UY is user interaction data and FY is user feedback data. and is a weight parameter used to control the impact of the data source on the score;

[0207] Based on the preliminary indicator set, compared with the quality standards, iteratively adjusted and refined the matching indicators, using the formula:

[0208] MY=ρ Y ·(IY+QY)

[0209] Get the content matching index MY, where QY is the quality control parameter, IY is the preliminary index set, ρ Y It is the adjustment coefficient used to balance the relationship between the initial indicator and the quality standard.

[0210] Assume the following values

[0211]

[0212] Uy=75;

[0213] FY=85.

[0214] Calculation process:

[0215]

[0216] It means that after comprehensive consideration of user interaction and feedback, the preliminary indicator score for content matching is 79, indicating high user engagement and satisfaction.

[0217] Assume data:

[0218] ρ Y =1.1,QY=5.

[0219] Calculation process:

[0220] MY=1.1·(79+5)

[0221] MY=1.1·84=92.4

[0222] It means that after considering the quality control parameters, the adjusted content matching index is 92.4, indicating that the matching degree has been improved through adjustment.

[0223] See also Figure 7 , the specific steps for analyzing user learning effects and behavior patterns are:

[0224] Based on the content matching index, the user interaction data is integrated to calculate the interaction score using the formula:

[0225] Score interact =Fw t ·T L +Fw c ·C L

[0226] Get the user interaction score Score interact , where T L is the residence time, C L is the number of clicks, Fw t and Fw c They are the weight coefficients of time and click respectively;

[0227] Based on the user interaction score, identify the learning behavior pattern, using the formula:

[0228]

[0229] The relationship between the interaction score and the learning outcome is expressed as Slope, where x i is the individual value of the interaction score, y i is the corresponding learning outcome, and is the average of the interaction score and the learning outcome.

[0230] Assume the following data:

[0231] T L 30 minutes;

[0232] C L 15 times;

[0233] FWf t =0.7;

[0234] FWf c =0.3.

[0235] The interaction score is calculated as follows:

[0236] Score interact =0.7·30+0.3·15=21+4.5=25.5

[0237] The score indicates the overall quality of user interaction with specific content, with higher scores indicating more active user interaction.

[0238] Suppose there are three data points:

[0239] (10,200), (15,240), (20,300)

[0240] First calculate the average

[0241] Then calculate the slope:

[0242] (10-15)(200-246.67)+(15-15)(240-246.67)

[0243]

[0244] Slope = 10 indicates that for every unit increase in interaction score, the learning outcome increases by 10 points on average.

[0245] See also Figure 8 , the specific steps for obtaining the refined content recommendation results are:

[0246] According to user interaction data, adjust the recommendation logic of research content, optimize the pertinence and accuracy, and analyze the relationship between user behavior and content preferences. Use the formula:

[0247]

[0248] Get the optimized recommendation logic, where GE x represents the user feature vector, θ is the parameter vector of the logistic regression model, P(y x =1|GE x ) is the given user feature GE xNext, the predicted probability of liking a certain content, θ T GE x is the dot product of the parameter vector and the eigenvector;

[0249] Based on the optimized recommendation logic, the recommendation score is calculated and the content ranking is determined using the formula:

[0250] DScore = α B PDY+β B ·RLC

[0251] The refined content recommendation result DScore is obtained, where PDY represents the popularity of the content, RLC is the correlation between the content and the user's preference, and α B and β B is a tuning parameter.

[0252] Assume there is a user feature vector GE x =[1,2] (for example, 1 represents the user's age group, and 2 represents the user's click history);

[0253] A model parameter vector θ = [0.5, -0.4];

[0254] Calculate θ T GE x The dot product of:

[0255] θ T GE x =0.5×1-0.4×2=0.5-0.8=-0.3

[0256] Then P(y x =1|GE x )for:

[0257]

[0258] It means that given these features, the model predicts that the user likes the content with a probability of about 42.6%.

[0259] Assume that the popularity of a piece of content is 200 (e.g. 200 hits), the relevance score is 0.8, and the weight parameter is set to α B = 0.5 and β B =1.5, the recommended score is calculated as:

[0260] DScore=0.5×200+1.5×0.8=100+1.2=101.2

[0261] The score will be used to compare this content with other content to determine which content is most appropriate to recommend to the user.

[0262] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. The elderly study decision-making system based on data analysis is characterized by: The system comprises: The learning behavior analysis module is based on the learning activity records of elderly users. Through cloud computing and edge computing, it collects learning time, content browsing and interaction frequency, creates corresponding timestamps for each activity, and records the attributes and time information of each learning activity to generate an overview of behavioral characteristics. The learning effect monitoring module uses data cleaning technology to remove abnormal records based on the behavioral feature overview, conducts linear analysis on the learning progress, quantitatively evaluates the user's understanding depth and memory retention rate, and generates quantitative analysis information on the effect; The study content adaptation module identifies matching learning topics and materials based on the quantitative analysis information of the effects, combined with user preferences and historical behaviors, automatically generates targeted study content through the NLP model, and formats the content into text, images or videos to obtain content matching indicators; The decision optimization support module analyzes the user's learning effects and behavior patterns based on the content matching indicators, optimizes the research and study content recommendation process, reduces the user's selection range, and matches the most suitable content options to obtain detailed content recommendation results.

2. The elderly study decision-making system based on data analysis according to claim 1 is characterized in that: The steps for obtaining the behavioral feature overview are specifically as follows: Based on the learning activity records of elderly users, the learning data of elderly users are collected through cloud computing and edge computing, including the length of learning time, content browsing volume and interaction frequency, using the formula: D i =(T i ,V i ,F i ,t i ) Get the data tuple D of the i-th activity i , where T i represents the learning time length of the ith activity, V i represents the number of content views of the i-th activity, F i represents the interaction frequency of the ith activity, t i Represents the timestamp of the i-th activity; Aggregate the data tuples and calculate the weighted average of the activities using the formula: and and Get the weighted average length of study time Content views and interaction frequency Among them, w T 、w V and w F Represent the weight parameters of learning time length, content browsing volume and interaction frequency, T i represents the learning time length of the ith activity, V i represents the number of content views of the i-th activity, F i represents the interaction frequency of the ith activity; Based on the weighted average learning time length, content browsing volume and interaction frequency, combined with the activity time attribute, an overview of the behavioral characteristics is obtained.

3. The elderly study decision-making system based on data analysis according to claim 1 is characterized in that: The steps of performing linear analysis on the learning progress are specifically as follows: Based on the behavioral characteristics overview, the cleaned learning data is integrated to extract each user's learning time and number of completed courses using the formula: Get the preliminary learning progress data PD, where: is the learning time of the ith data point, is the corresponding number of completed courses; Based on the preliminary learning progress data, a linear regression analysis is performed to calculate the best fitting straight line parameters using the formula: and Get the linear model parameters, where a d is the slope, b d is the intercept, n d is the total number of data points, is the learning time of the ith data point, is the corresponding number of completed courses; Using the linear model parameters, the learning progress of future or unfinished data is predicted using the formula: c predicted =a d ·t d +b d Get the predicted learning progress result, where a d is the slope, b d is the intercept, c predicted It is based on the user's future learning time t d The predicted number of course completions.

4. The elderly study decision-making system based on data analysis according to claim 1 is characterized in that: The steps for obtaining the effect quantitative analysis information are specifically as follows: Based on the learning progress information, the predicted learning outcomes of each user are calculated, and the estimated values ​​of the understanding depth and memory retention rate of each user are extracted to obtain a preliminary evaluation result set; Based on the preliminary evaluation result set, the prediction results of each user are analyzed using the formula: Calculate the depth of understanding and memory retention rate to obtain the quantitative evaluation result QR and obtain the quantitative analysis information of the effect, where R s is the user's predicted learning outcome, α r is the sensitivity coefficient, γ r is the linear influence coefficient, δ r is the smoothing parameter.

5. The elderly study decision-making system based on data analysis according to claim 1 is characterized in that: The steps for identifying the learning topics and materials are specifically as follows: Based on the quantitative analysis information of the effect, the learning outcomes of each user are extracted, and the adaptability score of each learning topic is calculated by combining the user preference data and historical behavior, using the formula: Get the user's topic matching score ES i , where EG i Represents the learning outcome score, EP i Indicates preference score, EH i represents the historical behavior score, a x , b x and c x is the weight parameter; The topic matching scores of the users are screened, and the learning topics with the best scores are selected. The NLP model is used to automatically extract the learning materials associated with the topics from the content library, using the formula: Get the matching learning material list LX, where NLP_score j is the relevance score given by the NLP model, is the weight coefficient of the material relevance score, n x is the material quantity; Based on the matching learning material list, the material selection is optimized to match the educational needs, using the formula: Get the list of learning topics and materials, where OX represents optimization score, Diversity_score k and Depth_score k are the diversity and depth scores of the materials, λ x and μ x is the influence that regulates diversity and depth, and m is the number of materials that have been screened and evaluated.

6. The elderly study decision-making system based on data analysis according to claim 1 is characterized in that: The steps for obtaining the content matching index are specifically as follows: Collect and organize user interaction data and user feedback data, initialize the calculation of matching indicators, and use the formula: Get the initial indicator set IY, where UY is user interaction data and FY is user feedback data. and is a weight parameter used to control the impact of the data source on the score; Based on the preliminary indicator set, the matching indicators are iteratively adjusted and refined by comparing with the quality standards, using the formula: MY=ρ Y ·(IY+QY) Get the content matching index MY, where QY is the quality control parameter, IY is the preliminary index set, ρ Y It is the adjustment coefficient used to balance the relationship between the initial indicator and the quality standard.

7. The elderly study decision-making system based on data analysis according to claim 1 is characterized in that: The analysis steps of the user learning effect and behavior pattern are specifically as follows: Based on the content matching index, the user interaction data is integrated to calculate the interaction score using the formula: Score interact =Fw t ·T L +Fw c ·C L Get the user interaction score Score interact , where T L is the residence time, C L is the number of clicks, Fw t and Fw c They are the weight coefficients of time and click respectively; Based on the user interaction scores, learning behavior patterns are identified using the formula: The relationship between the interaction score and the learning outcome is expressed as Slope, where xi is the individual value of the interaction score and y is i is the corresponding learning outcome, and is the average of the interaction score and the learning outcome.

8. The elderly study decision-making system based on data analysis according to claim 1 is characterized by: The steps for obtaining the refined content recommendation results are specifically as follows: According to user interaction data, adjust the recommendation logic of research content, optimize the pertinence and accuracy, and analyze the relationship between user behavior and content preferences. Use the formula: Get the optimized recommendation logic, where GE x represents the user feature vector, θ is the parameter vector of the logistic regression model, P(y x =1|GE x ) is the given user feature GE x Next, the predicted probability of liking a certain content, θ T GE x is the dot product of the parameter vector and the eigenvector; Based on the optimized recommendation logic, the recommendation score is calculated and the content ranking is determined using the formula: DScore=a B ·PDY+β B ·RLC The refined content recommendation result DScore is obtained, where PDY represents the popularity of the content, RLC is the correlation between the content and the user's preference, and α B and β B is a tuning parameter.

Images