Projects

Sentiment Analysis

🧠 NLP Sentiment Rating Prediction Project — Key Highlights

📌 Project Goal: Built an NLP model to predict sentiment ratings (0–10) from textual feedback in a coaching institute context.
🔍 Data: 200,000+ unique reviews generated with realistic coaching-related expressions (e.g., "teachers good", "not up to mark", "average", etc.).
🧹 Preprocessing: Applied text cleaning, tokenization, stopword removal, and lemmatization using NLTK and re.
🧠 Feature Engineering: Used TF-IDF vectorization and Word2Vec to convert text into numerical features.
📈 Modeling: Trained regression models including Linear Regression to predict sentiment scores.
🧪 Evaluation: Achieved high prediction accuracy with reduced MAE and RMSE; selected the best-performing model for deployment.
🚀 Deployment: Deployed the final model using Streamlit for real-time prediction with a user-friendly web interface.
💼 Business Impact: Automated sentiment scoring reduced manual review time, minimized human bias, and improved feedback analysis efficiency by 74%

Try it out!!

Objective:

To analyze one year of realistic, multi-source e-commerce data and deliver data-backed strategies to achieve at least 📈 15% revenue growth through optimization across marketing, logistics, and customer experience.

📊 Defined and monitored KPIs: Net Revenue, Profit Margin, AOV, Return Rate, Cart Abandonment, CAC, ROAS
🧠 Built Excel and Power BI dashboards to explore trends across loyalty, product category, city tier, and conversion funnel
🛠️ Used SQL (BigQuery) for deep-dive analysis using joins, aggregations, and window functions
🧬 Applied Cohort Analysis & RFM Segmentation in Python to identify customer lifecycle groups (e.g., Champions, At-Risk)
📐 Conducted statistical testing (T-Test, ANOVA, Chi-square, Z-Test) to validate differences with statistical confidence
🔁 Simulated 9+ business strategies (e.g., loyalty conversion, AOV uplift, funnel improvement) to forecast revenue impact
🌐 Developed a clean, interactive Streamlit app for real-time KPI tracking and scenario simulation

🧰 Tools Used:

✅ Outcome:

Achieved a 15.03% increase in projected annual revenue through strategic optimization, simulation, and data storytelling — all backed by domain knowledge and statistical evidence.

Blinkit Revenue Optimization & Simulation Dashboard

Try it out!!

Student Churn Prediction Engine

🎯 Objective

To proactively identify students at risk of dropping out from a coaching institute and enable timely interventions to improve retention and reduce revenue loss.

🛠️ Tools & Technologies

Python, pandas, scikit-learn, Streamlit, pickle (for model deployment)

📊 Project Overview

I developed a machine learning model using Random Forest to predict student churn based on academic performance, attendance, feedback sentiment, and behavioral scores. The final model achieved 97% accuracy and was deployed as an interactive Streamlit web app for real-time prediction and counselor use.

💼 Business Impact

Correctly flagged 14,142 high-risk students
Enabled interventions that retained 5,656 students
Saved approximately ₹2.83 Crores in potential revenue loss

🚀 Outcome

This project showcases my ability to solve real business problems using end-to-end data science — from feature engineering and model building to deployment and impact simulation.

Try it out!!