IPO AI — Indian IPO Predictor

Machine learning predictions on Indian IPO listing performance using GMP, subscriptions, and market data.

Next.jsPythonXGBoostPostgreSQLPrismaDocker

ipo-ai.app

Overview

Indian retail investors had no data-driven tool to evaluate IPO subscription decisions. Decisions were made on social media hype and broker tips rather than quantitative signals. This system brings ML-grade analysis to a previously intuition-driven market.

The Problem

The Indian IPO market sees dozens of listings per quarter. Retail investors were making decisions based on grey market premiums and promoter reputation alone, missing the compounding signals that predict listing outcomes.

The Solution

A scraping pipeline collects historical IPO data — subscription rates across investor categories, GMP history, and index data. XGBoost classifiers and regressors predict listing day performance. The Next.js frontend surfaces predictions with confidence scores.

Key Features

XGBoost Prediction Models

Separate classification (above/below issue price) and regression (listing gain %) models trained on 300+ historical IPOs.

Auto-Scraping Pipeline

Scripts run on schedule to collect live subscription data, GMP updates, and Nifty/VIX index values.

GMP Signal Integration

Grey market premium data ingested and normalized as a key feature, significantly improving model accuracy.

Accuracy Tracking

The site tracks model prediction accuracy against actual listing results over time — full transparency on how the model performs.