Skip to main content

Welcome to Lead Scoring Model

The Lead Scoring Model is a machine learning solution that predicts the probability of lead conversion for Event Management SaaS applications. Built with Gradient Boosting, this model achieves 90.4% accuracy in classifying leads into conversion outcomes.

Key Capabilities

High Accuracy Predictions

Gradient Boosting model achieving 90.4% accuracy with 0.91 cross-validation score

Multi-Model Comparison

Compares 12 classification algorithms including Random Forest, AdaBoost, SVM, and Neural Networks

Advanced Data Preprocessing

Automated data fusion, missing value imputation, label encoding, and feature scaling

Production Ready

Integrated with Shimoku API for dashboard visualization and real-time predictions

How It Works

The model processes two primary datasets:
  1. Leads Data - Information about all potential clients including source, use case, acquisition campaign, and demographic details
  2. Offers Data - Details about clients who reached the demo meeting stage, including pricing, discount codes, pain points, and conversion status
Through intelligent data fusion and feature engineering, the model identifies patterns that predict whether a lead will convert to a paying customer.

Model Architecture

The system evaluates 12 different classification algorithms:
  • Ensemble Methods: Random Forest, AdaBoost, Extra Trees, Bagging Classifier, Gradient Boosting
  • Tree-based: Decision Tree
  • Probabilistic: Naive Bayes
  • Instance-based: K-Nearest Neighbors
  • Linear Models: Logistic Regression, SGD Classifier
  • Neural Networks: Multi-Layer Perceptron
  • Support Vector Machines: SVM
The Gradient Boosting Classifier emerged as the best performer with a cross-validation score of 0.91.

Target Classification

The model classifies leads into three categories:
  • Closed Won - Leads that converted to paying customers
  • Closed Lost - Leads that did not convert
  • Other - Leads in intermediate states (grouped from minority classes to handle class imbalance)

Get Started

Installation

Set up the project and install dependencies

Quickstart

Train your first model and make predictions

Build docs developers (and LLMs) love