Fuzzy rough sets theory-web and mobile based recommendation system for students career

IN594976BActive Publication Date: 2026-07-10DR VE JAYANTHI +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
IN202041018119
Authority / Receiving Office
IN · IN
Patent Type
Patents
Current Assignee / Owner
Filing Date
2020-04-28
Publication Date
2026-07-10
Estimated Expiration
2040-04-28

AI Technical Summary

Technical Problem

Conventional recommendation systems face issues with data sparsity and prediction accuracy, particularly when recommending careers to students based on their educational and skill data, as they rely on traditional algorithms that do not effectively capture the nuances of students' mental abilities and knowledge levels.

Method used

The development of a novel algorithm, Fuzzy Rough Set Theory Based Collaborative Filtering (FRSTBCF), which combines with Content-Based Recommendation (CB) to enhance accuracy by using indiscernibility relations and fuzzy T-equivalence to predict student proficiency in subjects, and is implemented in both web and mobile platforms for career guidance.

Benefits of technology

The FRSTBCF algorithm improves prediction accuracy and precision compared to traditional methods, providing better career recommendations by effectively assessing student knowledge levels and subject interests, with lower error rates and improved user engagement through mobile accessibility.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present disclosure is directed to methods and systems for identifying career recommendation system based on education domain. Recommendation Systems (RS) is capable to perform an eminent task in the internet world; Collaborative Filtering (CF) is one of the popular methods in RS. At the maximum, RS is implemented in a business intention to recommend the products, movies, travel guidance etc. But in this work, the efficient advisory tool is developed for the younger generation to choose their right career based on their knowledge level. Finding the student's mental ability and the acquired knowledge from their given data is the major aspiration of this advisory tool. RS suffers the data sparsity problem, recommendation in accuracy and big error in predictions etc., CF based RS is used to prevent the data sparsity issue and less accuracy. So, a hybrid FRST-IBCF-CB technique is proposed for reducing the big error in RS prediction. Web based FRST-IBCF-CB technique is used to develop the student career recommendation system. Here, cross validation approach is considered for segregate the dataset into training set and test set of the common repository data. Root Mean Square Error (RMSE), Mean Square Error (MSE) and Mean Absolute Error (MAE) values are determined to measure the rating prediction accuracy. To evaluate the entire students’ career recommendation system, real time data is collected from engineering students. With the help of that student’s feedback, the metrics like true positive, false positive, true negative and false negative are measured to find the correctness of the proposed model. Accuracy of the students’ career recommendation system is 85% and career wise classification precision values are more than 80%. Android OS based Mobile app is developed for the student's career recommendation system. To achieving this, the packages of the R Language "shiny" and "recommenderlab" is used to build the core part of the proposed model. The Android's component WebView is used to fulfill the Mobile App development
Need to check novelty before this filing date? Find Prior Art

Description

COMPLETE SPECIFICATION(See section 10 and rule 13)1. TITLE OF THE INVENTION: Fuzzy Rough Sets Theory-Web and mobile based Recommendation System for Students Career 2. APPLICANT(S)Name Nationality AddressDr. VE. Jayanthi Indian Professor & Head, Department of Electronics and Communication Engineering, PSNA college of Engineering and Technology, Dindigul, Tamil Nadu India 624622C. Raja Kumar Indian 59, Pettai Road, Begambur, Dindigul, Tamilnadu India 624002PREAMBLE TO THE DESCRIPTION PROVISIONALThe following specification describes the invention. COMPLETEThe following specification particularly describes the invention and the manner in which it is to be performed.4. DESCRIPTION (Description shall start from next page.)5. CLAIMS (not applicable for provisional specification. Claims should start with the preamble— “I / We claim” on separate page)6. DATE AND SIGNATURE (to be given at the end of last page of specification)7. ABSTRACT OF THE INVENTION (to be given along with complete specification on separate page) Field of Invention This present invention is directed to methods and systems for identifying career recommendation system based on education domain via internet based platform using web based interface as well as mobile application.Back ground of the invention Most of the recommendation systems are related to social networking. Generally, Employer career system is developed and they used the input as resume, digital document and personal detail. With this above data the system is suggested their carrier decision. Developing the recommendation system they use input as user's information like educational details, personal details, employment details, skills, position histories, salary histories, organization data, etc., the information is collected in the form of the digital resume or receives that information through the web user interface. This recommendation received from those systems is a career path for the employee, a career-related event, career suggestions, salary suggestions, career path within an organization based on different algorithms.Summary of the InventionRecommendation Systems (RS) is capable to perform an eminent task in the internet world, Collaborative Filtering (CF) is one of the popular methodology in RS. It suffers from the data sparsity problem, recommendation inaccuracy, big-error in predictions etc., To enhance the accuracy of the RS and reducing the big-error in prediction, a novel algorithm named as Fuzzy Rough Set Theory Based Collaborative Filtering (FRSTBCF) is developed. At the maximum, the RS is implemented in a business intention to recommend the products, movies, travel guidance etc. But in this work, the efficient advisory tool is implemented for the younger generation to choose their right career based on their knowledge level. Finding the student's mental ability and the acquired knowledge from their given data is the major aspiration of this tool. For this fusing the FRSTBCF algorithm with the traditional Content-Based Recommendation (CB) is determined and implemented. These days, everyone is expecting their day to day needs and the way to access it to be there in their hands. Hence, a mobile app for the student's career recommendation system is taken into consideration. This is a great potential to connect and collaborate with each students. It reinforces the brand value of the proposed RS too. A novel way is used to make the mobile app for the student's career recommendation system using the FRST-IBCF-CB technique. To evaluate the proposed model, data is prepared using the cross validation method. Based on that, ratings are evaluated by calculating the MAE (mean average error), MSE (means squared error) and RMSE (root means squared error) values. Further the correctness of the model is measured by finding rates like Accuracy, Specificity, Sensitivity, and Precision & False Positive Rate. The proposed FRSTBCF algorithm is compared with the traditional algorithms experiment results such as Item Based Collaborative Filtering using the cosine similarity (IBCF-COS),IBCF using the Pearson correlation (IBCF-COR), IBCF using the Jaccard similarity (IBCF-JAC) and Singular Value Decomposition approximation (SVD). The proposed algorithm gives better error rate and its precision value is comparatively identical with the existing system. Brief Description of the DrawingsRepresentations of the present invention are described in detail below with reference to the drawings wherein:Figure 1. An Overview of the proposed hybrid FRST-IBCF-CB technique.Figure 2. Web based User Interface.Figure 3. Collected Data in Repository.Figure 4. Collected Data in CSV Format.Figure 5. Familiarity Matrix (FM).Figure 6. Familiarity matrix of the ten studentsFigure 7. Mobile app deployment ArchitectureFigure 8. Mobile App Startup Flow ChartFigure 9. Mobile App & Repository Communication Flow ChartFigure 10. Mobile App & Recommendation System Flow ChartDetailed Description of the InventionMethodology In the conventional RS, similarities between the components are identified by using the cosine similarity measures (or) Pearson similarity measurements. But the FRSTIBCF algorithm acquires the notions of indiscernible relation from Fuzzy Rough Sets Theory (FRST). The indiscernibility relation is calculated to identify the similarity between the studied subjects, studying subjects by the students in an educational institution. It determines the degree to which those subjects are indiscernibility. The fuzzy T-equivalence relation with the equation Gaussian kernel function is considered to estimate the indiscernibility relation. From that, further process will takes place to obtain the recommendation. Once the similar subjects are found then rating for the new subject is predicted and the recommendation is given to the students. Prediction is done by using the traditional weighted sum techniquePeriodically, Universities / Institutions introduced more subjects for the enrichment of the student’s knowledge. A progressive methodology is taken in this proposed model which helps the students studying in those Universities / Institutions. Figure 1 shows an overview of the career recommendation system. The first three steps are needed for analyzing the CF and CB methods. Data normalization is the crucial routine in this methodology to standardize the final outcome. FRSTIBCF method is assisting the students to trace out the relevant subjects for improving their subject’s skill is explained. In addition to FRSTIBCF method, the proposed model found the students knowledge level in the unrated subjects and with the help of this their profiles are constructed. The generated profiles are the used to develop the student career RS.Acquiring the existing subject knowledge from studentsA unique web user interface with the well crafted, organized questionnaire is provided to the students community for collecting data. Figure 2 illustrate the web user interface. It is designed for the engineering branches like Biomedical Engineering, Computer science and Engineering &Electronics and Communication Engineering. It is used to collect the student’s knowledge level in their respective branches explicitly. The input received from the students is very much important for further process. Hence the questionnaire is designed in an easy to answerable way and formulated as a most interesting questionnaire with the respectful output. Generally the RS received the user input through the rating scale. The prediction accuracy of the RS depends on the provided input. To afford the accurate prediction at the end, the proposed model provides the graphical widget with decimal intervals. Hence students are able to give their inputs with fractional values.The Anna University regulation 2017 curriculum is considered in this user interface (UI). The eight different subjects in a particular engineering branch suggested by the Subject Matter Experts (SMEs) are prearranged in this questionnaire. The career outcome based subjects and its unit wise five questions are deliberated in this questionnaire. An exclusive framework is followed in the questionnaires. As like in Figure 2, the questionnaire for the computer science and engineering branch consist the subjects like C, Python, Data Structures, DBMS, OS, Networking, Java and Web Technology. In this, first students should select their appropriate branches, and then reveal their knowledge level for each and every question available in the questionnaire. All those unique questions have its own rating scale from 0 to 5 with the 0.5 interval. Some of the students may like the subject Si while other might like the subject Sj. Based on that, the students cited their proficiency levels in their appropriate branches; it is stored in a common repository available in the backend. The repositories are tabulated for each and every branch. The data collected from the computer science and engineering (CSE) branch students are stored in cse_repo table and it is shown in Figure 3.Normalizing the data in subject levelStudent may inflate their knowledge level or undervalue it while giving their ratings through the provided user interface. It is a quite common scenario, some of the student’s normal tendency too. To address this kind of the challenges, the data normalizing process takes before build the knowledge matrix. For that, the collected data is pulled out from the common repository as a Comma Separated Value (CSV) file. Figure 4 illustrates such kind of data related to the computer science and engineering branch students.Then the cited unit wise proficiency in their subjects by the students is normalized with weighted vector. At present, the proposed model configured the total weight for a subject as 20. But, the unit-wise weights for each and every subject in the respective branches are suggested by the SMEs. The unit wise weights given by SMEs in the CSE branch for the subject ‘C programming’ are represented in the Table 1. Usually, the weights for the units & the weighted vector for each and every subjects may vary.Table 1. Unit-wise weights for a subject Unit I Unit II Unit III Unit IV Unit VC Programming 3 3 4 5 5The average weight vector for the same subject is listed in Table 2. During a given proficiency level of a student in that particular subject is listed in Table 3.Table 2. Unit-wise average weight vector for a subject Unit I Unit II Unit III Unit IV Unit VC Programming 0.15 0.15 0.20 0.25 0.25Table 3. Unit-wise proficiency cited by a student Unit I Unit II Unit III Unit IV Unit VC Programming 0.5 0.5 1.0 1.5 0.5The values from the Table 2 are known as weight factor which is multiplied with the values in the Table 3 to do for normalizing the students input data.Finding the students / subjects familiarity matrixKnowledge level values are given by the students for the five different questions based on the subject is shown in Figure 2. This prediction is used to identify the legitimate knowledge level of the subject. Weighted average calculation uses the weight vector of the identified legitimate knowledge level in a particular subject and it is shown in Figure. 5. Subjects rating are denoted as Si and the familiarity matrix is denoted as FM. In subject rating matrix row represents the number of students and the columns represent the unit number in the subject is in equation 1. Using equation 2 the knowledge matrix KM is build. The summation of the cited proficiency level is multiplied by the unit-wise average weight and then it is divided by the sum of all weights. From this KM matrix, the students and their subjects’ knowledge are identified. For example, the values of subject S1 and cited value U1 mentioned in the Table 2 and Table 3 is substituted in equation 2 and its normalized value is 0.85. In similar way all the cited values against the subjects are identified and formed as a knowledge matrix. The knowledge matrix KM is a pair-wise comparison square matrix is shown in equation 3, where each element represent the relative importance of certain attribute am compared with another criterion an. In equation 3, 'n' represents the number of subjects and 'm' represents the number of students. Figure 6, illustrate the representation of the knowledge level matrix. The derived proficiency of the subjects is in rows. Predicting the unrated subject knowledge using FRSTStudents are not giving their proficiency level to all the subjects because those have not been studied subject they might not be able to give their citation (value). In this case FRSTIBCF algorithm is used to predict the subjects’ proficiency level (cited value) for the students. The values are in between 0 to 5. Similarity learning is the key mechanism takes place with the method T-equivalence and its fuzzy indiscernibility relations to find the students knowledge level in those subjects. Eight different subjects level knowledge (0 to 5) of the two students is listed in Table 4. Table 4. Subject wise students’ knowledgeSubject Student I Student II1 3.375 2.1752 2.075 3.053 2.2 1.84 1.875 3.2755 0 1.756 1.425 2.7257 1.75 1.958 1.4 2.525A view of the prediction from that Table 4 is, the student1 is skilled in subject1 rather than other subjects. Similar to that, the student2 is very much skilled in subject4 comparably with other subjects. Ordering the subject knowledge gives an idea about the subject interest of the student which is needed for finding their carrier.Building the students’ profileAll the subjects in a program offered by the Anna University is coming under various characteristic like programming, algorithm, concepts, modeling, design, analytics and tools. A binary illustration of the subject characteristics are tabulated in Table 5. For example, the subject1 is very much related to programming, algorithm and design not about modeling, concepts and analysis and it is used to construct the proposed model to identify the students’ skill.Table 5. Characteristics of the individual subjectsSubjects Programming Algorithm Design Model Concepts Analysis1 1 1 1 0 0 02 0 0 0 1 1 13 0 0 0 1 1 04 0 0 1 1 1 15 0 1 0 0 1 16 1 0 0 1 1 07 1 1 1 0 0 18 1 0 1 0 1 0The information retrieval system is uses the concept of Term Frequency (TF) and Inverse Document Frequency (IDF) and it is discussed

[14] . The same is applied to discover the relationship between the subjects available in the curriculum and the students. The TF / IDF is used to analyze the characteristic weight of the particular subject using equation 4. TF is used to determine the relative characteristics of the subject and it helps to identify about how often the specific characteristic of the subjects are looking about is appear in the subject. It is calculated by dividing the term occurrence by the square root the number of characteristics in a subject. The student’s liking is identified based on their studied subject knowledge level. With the help of the student’s liking and the studied subject’s outcome, the career is advised to the students.Table 6. Subjects characteristics and its Term Frequency (TF) / Inverse Document Frequency (IDF).Sub. Programming Algorithm Design Model Concepts Analysis Total no. of attributes1 1 1 1 0 0 0 32 0 0 0 1 1 1 33 0 0 0 1 1 0 24 0 0 1 1 1 1 45 0 1 0 0 1 1 36 1 0 0 1 1 0 37 1 1 1 0 0 1 48 1 0 1 0 1 0 3 DF 4 3 4 4 6 4 IDF 0.397940009 0.522878745 0.397940009 0.397940009 0.22184875 0.397940009 Since the subject’s characteristics are represented in a binary format, normalization is preceded to calculate the TF. The binary representation of the subjects and its characteristics are normalized by dividing the characteristics occurrence by the square root of number of attributes in the subject. Hence, for Subject1, the normalized value is calculated like = 0.577350269. Table 7 illustrates the normalized value of the subject characteristics.Table 7. Normalized value of the subject characteristicsSubjects Programming Algorithm Design Model Concepts Analysis1 0.5773503 0.5773503 0.5773503 0.0000000 0.0000000 0.00000002 0.0000000 0.0000000 0.0000000 0.5773503 0.5773503 0.57735033 0.0000000 0.0000000 0.0000000 0.7071068 0.7071068 0.00000004 0.0000000 0.0000000 0.5000000 0.5000000 0.5000000 0.50000005 0.0000000 0.5773503 0.0000000 0.0000000 0.5773503 0.57735036 0.5773503 0.0000000 0.0000000 0.5773503 0.5773503 0.00000007 0.5000000 0.5000000 0.5000000 0.0000000 0.0000000 0.50000008 0.5773503 0.0000000 0.5773503 0.0000000 0.5773503 0.0000000The very important task to deal with the CB technique is to find the relevant attributes. This is used in the proposed model to regulate the subjects and gives its meaningful recommendation. In the literature two approaches are represented in content based recommendation technique. One of the approaches is just analyzing the description of the content. Another approach facilitates to find the student view and the subjects profile. The second approach is considered in the proposed model.In the students profile building process, calculate the sum of the product by multiplying each normalized characteristics column with each student’s knowledge level values appropriate to the subjects. Table 8 shows the students profile after calculating the sum of products value. From that, it clearly articulate the students view.Table 8. Students ProfileStudents Programming Algorithm Design Model Concepts Analysis1 4.454572 2.823557 0.5773503 4.513861 5.322151 3.0105022 5.261826 3.241100 5.326046 6.244490 8.712662 5.383781The student1 is very much interested to learn the concepts based subjects followed by the subjects related to the design. Student2 is very much likes to learn the subject related concepts and then model based subjects. The sorted profile for the student2 is displayed below.Table 9. Sorted Students ProfileProgramming Algorithm Design Model Concepts Analysis8.712662 6.244490 5.383781 5.326046 5.261826 3.241100 Career RecommendationThe subjects rating are predicted based on the available values identified by the proposed model. For example, Table 10 is derived for the dot product of the IDF value from Table 9, normalized subjects characteristics values from Table 7 and the students profile value from Table 8.Table 10. Subjects rating based on students interestSubject Student I Student II1 2.925640 3.4110092 2.410413 3.7875613 2.105027 3.1238744 2.996642 4.3398475 2.225737 3.3313216 2.742189 3.7595417 3.132678 4.0252318 2.754937 3.548528Table 10 illustrates the student1 interested to learn the subject7 than subject4 and subject1 and so on. Based on the input given by the SMEs the subject characteristics are categorized into 3 groups. The groups are named like industry, research and higher studies. The characteristics like programming and algorithm are considered for the category of industry. The concepts and the model characteristics are under the research category and then the design and analysis are in higher studies category. To find the career suggestion, the proposed system model sorted top 3 characteristics from the students profile and intersects with its corresponding categories.The proposed model's logics are executed in the web server which is not able to execute in the mobile or handheld devices. Android based app's gives access to all types of devices like smart phone, tablet PC, TVs etc., Mobile app gives the provision to send the input to the backend server and render the response back from the server to the end user.Proposed model based upon Shiny appThe R package Shiny is used to build the proposed model's UI. It provides easier steps to build interactive and responsive web applications (apps) straight from R language. It can be built entirely using R language, or R language combined with HTML, CSS and JavaScript. In general, the logical server side script for the Shiny app is cleanly separated from the UI design part. The logical script is written in the file server.R and the UI design script is written in ui.R file. Various widgets like sliderInput, textInput, fileInput etc are available in shiny package. These widgets are scripted in the ui.R file. After receiving the input from the user, the server side script performs the machine learning process and returns the output. Reactive / responsive of the server side script is in the form of textOutput, plotOutput, tableOutput, uiOuput etc. Shiny web apps require a web server to execute.Hosting the proposed Shiny app on the Shiny serverTo run the proposed model it should be hosted on the web server. The server is a local server provided by RStudio or the existing server provided by shinyapps.io. The architecture deployment setup is illustrated in Figure 7.Develop Android appTo build an Android app, the Java Development Kit (JDK) and the Android Studio IDE are commonly used. An Android apps is a combination of components like Activity, View, Intent, Services etc., The View component is a user interface element and it is also termed as Android Widgets. Android platform provides a collection of prebuilt widgets. The app users are able to view and interact with it. Some of the common widgets are Button, TextView, EditText, ListView, CheckBox, RadioButton, Spinner and many more.The WebView is a Java class, it facilitates to deliver a web application as a part of android app. The proposed model's web URL is configured with this class. Using the Android's SDK tools the developed Java source codes are compiled, packed with the resource files and make it as an APK file. Any Android OS devices are able to install the proposed app from the APK file.Execution Procedure The proposed model consists of two processes. The students facing front end process is the Android mobile based app, and the next process is the Shiny based web application which doing the real business logic for the career recommendation. The execution flow of the proposed model is as follows. The flow chart seen on Figure 8, 9 and 10 gives an overview about an end to end execution process.1. Initially, the students launched the proposed mobile app from their Android mobile phones or Android based handheld devices.2. The Mobile app gives the provision to the students for selecting their course branches. As a prototype, the proposed Mobile app shows the engineering branches like BME, CSE, ECE only. It has an option called more, to choose other branches.3. Based on the selected branch the appropriate questionnaire UI is displayed in the Mobile App. If the branch selection is not correct, then the error message "select your appropriate branch" is displayed in the mobile screen4. The given familiarity in the subject by the students are collected by the Mobile App and stored in to the common repository.5. Additionally, students are able to do the steps 2 and 3 through the web browser in their computer / laptop.6. The stored data is normalized for further process by the backend shiny web application.7. The shiny web application builds the Student / Subject familiarity matrix8. The unrated subject familiarity is predicted by the FRST-IBCF technique9. Based on the predicted values, the students profiles building process is executed10. Career for the students are identified and recommended.11. The identified recommendation is displayed in the Student's Mobile screen

Claims

Claims:We Claim: 1) Web based FRST-IBCF-CB technique is used to develop the student career recommendation system. 2) The technique in claim 1, where in the system collects the real time data from students. Questionnaire is based on syllabus and the students mark their subject knowledge in the form of floating point. This is the input of our developed system. 3) The system of claim 1 is able to access from the web browser and Android Mobile App. , Description:FORM 2THE PATENTS ACT, 1970(39 of 1970)&THE PATENTS RULES, 2003