PhonePredict - Mobile Price Prediction

Full-stack platform for AI-powered mobile price prediction and market comparison.

Overview

PhonePredict is a full-stack machine-learning platform that turns phone specifications into an estimated fair price and then compares that estimate with phones currently available in the market.

The prediction model is trained on approximately 28,000 phone configurations using K-Nearest Neighbours (KNN). After generating an estimate, the platform uses live market search and AI-powered structuring to surface relevant phones for comparison.

Highlights

Built the platform with React and FastAPI, providing a responsive frontend and lightweight API layer.
Trained a KNN machine-learning model on approximately 28,000 phone records, using around 8 neighbours to estimate fair prices.
Designed a simple prediction flow based on practical phone specifications including brand, rating, RAM, and storage.
Added live market search so predicted prices can be grounded against phones currently available for comparison.
Used AI to transform raw market results into five structured recommendations with names, prices, specifications, and relevance.

ML & AI focus

The prediction pipeline combines traditional machine learning with live market intelligence. The KNN model provides the initial price estimate, while AI structures and summarizes current search results into clear, comparable options.

This keeps the predicted value separate from actual listing prices while giving users useful market context around the estimate.

Architecture

The application follows a full-stack architecture with React handling the user experience and FastAPI powering the backend and prediction workflow.

The ML model is integrated into the backend so specifications can be processed consistently, predictions can be generated through an API, and the resulting budget can be passed into the market-search pipeline.

DevOps

The application is Dockerized to keep the development and deployment environment reproducible. The project is designed as a complete full-stack application, combining the frontend, backend, machine-learning model, and market-search workflow into a single product.