About project

Category: Machine Learning | Sales Analytics | Predictive Modeling

Status: Completed

Tech Stack: Python, Pandas, NumPy, Scikit-learn, XGBoost, LightGBM, Matplotlib, Jupyter Notebook

Dataset: deal_details.xlsx (CRM sales activity and deal attributes)

Target: Deal Outcome (Won or Lost)

GitHub: https://github.com/Nishigandha-Wankhade/Deal_Closure_Prediction_Using_ML/tree/main

Project Overview

Built and compared multiple machine learning classification models to predict whether a sales deal will be Won or Lost, using CRM activity signals and deal characteristics. The goal was to help sales teams prioritize high-probability deals, improve conversion rates, and strengthen revenue forecasting.

Problem Statement

Sales teams manage many open deals at once, but not all deals deserve equal attention. Without a data-driven approach, effort can be wasted on deals unlikely to close.

This project addresses that problem by predicting deal outcomes based on historical patterns in:

• Email, calls, meetings, tasks, notes, conversations
• Deal size, deal type, close date related signals

Dataset Features

Key fields used in modeling:

• id: deal identifier
• email_count, call_count, task_count, meeting_count
• notes_count, conversation_count
• deal_size