Student Performance Prediction and Early Warning System

About the project

Student performance prediction Android project that warns about weak students weeks before the exam. In most colleges low attendance and poor unit test marks are noticed only after the result. SafalSanket collects attendance and marks from the mentor's phone, builds seven inputs for every student and runs a trained neural network inside the app. Every prediction comes with a risk score and simple reasons like "Attendance 61%" or "Unit tests 33%", so the mentor knows exactly what to fix.

How it works

• The mentor marks date-wise attendance and enters Unit Test, Assignment and Prelim marks (CSV import supported)
• The server works out 7 inputs per student: attendance %, test %, assignments %, previous SGPA, backlogs, study habit score and marks trend
• One tap runs the neural network for the whole class on the phone, with no internet or cloud needed
• At-risk students are listed by risk score with the reasons behind each prediction
• The mentor adds an intervention (counselling, remedial class, parent meeting) and can alert the parent by SMS
• Students and parents see the same prediction, try the what-if planner and follow mentor remarks
• The admin sees a college-wide early warning dashboard and prints reports

Users

• Mentor (app): attendance, marks, class prediction, at-risk list, interventions, parent alerts, class report PDF
• Student / Parent (app): my prediction, what-if planner, study habit check, records, remarks, goals and streak, progress report PDF
• Admin (web panel): departments, classes, subjects, mentor approval, students, prediction records, reports, notices, settings

Highlights

• 3 roles, 22 Android screens, 14 admin pages, 33 APIs, 14 database tables
• Neural network (7-16-12-4) trained on 6,000 records: 86% accuracy, 93.9% accuracy for the "needs help" decision
• Explainable prediction: each input shows how much it raises or lowers the risk
• What-if planner: "attend 20 more lectures and score 55% → Pass"
• Validation on the phone and again in PHP: email, 10-digit mobile, marks ≤ max, no future attendance date
• Secure: password_hash, PDO prepared statements, role and class ownership checks
• Sample data and demo logins ready for the viva

What you get

• Complete Android Studio project (Java) and PHP source code
• MySQL database file with sample data
• Python dataset and training scripts with charts (confusion matrix, feature importance)
• Project report (Word, 50 pages) with DFD, ER and all UML diagrams
• 2-page project abstract (Word + PDF)
• Project explanation with 20 viva questions
• Setup guide (PDF) and installation help

Who can use this project

BE / B.Tech Computer, IT and AI & DS students, MCA, BCA and Diploma students looking for a machine learning Android project with source code, report and documentation. It fits well as an educational data mining or early warning system project.

Frequently asked questions

Does the prediction need internet? No. The model runs inside the app; internet is needed only to load and save data from the college server. Can I train the model again? Yes. The Python scripts are included; run train_model.py and copy the new JSON file into the app. Can a parent use the app? Yes. The parent logs in with the student's account and also gets alerts through the mentor's SMS. Which software is needed? Android Studio, XAMPP (PHP + MySQL) and an Android phone or emulator.