Customer-support-optimization-using-Voice-and-Non-Voice-Analytics-

🎧 Customer Support Optimization using Voice & Non-Voice Analytics

A full-stack analytics solution designed to enhance customer support operations across Email, Voice, and Chat channels. This project empowers support teams, operations managers, and CX strategists with predictive modeling and interactive dashboards to improve resolution efficiency and customer satisfaction.

🚗 GitHub Project Repository

🔗 Click to view Customer-support-optimization-using-Voice-and-Non-Voice-Analytics

🧠 Project Overview

Customer support is a critical touchpoint for retention and brand loyalty. This project analyzes 200,000+ support tickets to uncover resolution patterns, sentiment dynamics, and agent performance across multiple channels.

Key Objectives:

Clean and preprocess multi-channel support data
Engineer features for resolution modeling and dashboarding
Build classification and regression models for resolution prediction
Deploy interactive dashboards for stakeholder decision-making

📁 Project Structure

File Name	Description
`support_data.csv`	Raw dataset with ticket-level support info
`cleaned_support_data.csv`	Preprocessed dataset with feature engineering
`support_resolution_model.pkl`	Trained model for predicting resolution status and time
`customersupport.sql`	SQL queries for data extraction and filtering
`sqlconnect.py`	Python script for SQL database connection
`app.py`	Streamlit app for dashboard deployment
`customer_support.ipynb`	Jupyter notebook with EDA, modeling, and insights
`customer_support_optimization.pbix`	Power BI dashboard visualizing support KPIs and trends

🧹 Data Preprocessing

Verified and cleaned ticket-level data
Encoded categorical features (channel, issue_type, sentiment)
Normalized response and resolution time variables
Removed outliers and ensured type consistency
Converted date column to datetime format for temporal analysis

📈 Exploratory Data Analysis

Resolution time trends across channels and issue types
Sentiment distribution by support channel
Agent-level performance metrics
Monthly ticket volume and resolution status trends
Correlation matrix of support features

🤖 Modeling Approach

Target Variables: resolution_status, resolution_time_min
Algorithms Used: Logistic Regression, Random Forest, XGBoost
Evaluation Metrics: Accuracy, F1 Score, MAE, RMSE
Feature Importance: channel, issue_type, response_time_min, sentiment