Election Voting Prediction Model

📌 Project Overview

This project aims to build machine learning models for predicting which political party a voter is likely to vote for, based on survey data. The dataset consists of 1,525 voters and 9 variables, collected as part of an exit poll by a leading news channel, CNBE. The predictions will help in estimating the overall election outcome in the surveyed regions.

📝 Problem Statement

You have been hired by CNBE to analyze voter behavior using a dataset with 9 features. The goal is to develop machine learning models that predict the party a voter will choose, helping CNBE project election results more accurately.

🎯 Project Objectives

• Build effective classification machine learning models.
• Conduct thorough exploratory data analysis (EDA).
• Evaluate models using appropriate performance metrics.
• Select the best-performing model based on evaluation criteria.

📂 Dataset Details

• Total Records: 1,525
• Variables:
  1. vote – Political party/candidate the voter chose.
  2. age – Respondent's age.
  3. economic.cond.national – Perceived national economic condition.
  4. economic.cond.household – Household economic condition perception.
  5. Blair – Opinion rating of Tony Blair.
  6. Hague – Opinion rating of William Hague.
  7. Europe – Stance on European Union issues.
  8. political.knowledge – Level of political knowledge.
• Data Source: The dataset consists of two tabs: "Data" and "Data Dictionary." Only the "Data" tab is used for analysis.

🔍 Data Preprocessing & EDA

• Data Cleaning: Checked for missing values and duplicate records.
• Univariate & Bivariate Analysis: Studied the distribution and relationships between variables.
• Outlier Treatment: Applied the Interquartile Range (IQR) method to handle outliers in economic variables.
• Feature Engineering: Removed redundant features and optimized the dataset.

📊 Model Building

• Train-Test Split: Split data into training and testing sets.
• Applied ML Models:
  • Logistic Regression
  • Decision Tree (with pruning)
  • Naïve Bayes (optimized using Grid Search)
  • K-Nearest Neighbors (KNN with K=3, 5, 7)
  • Bagging & Boosting (GB Boost, XGBoost)
• Model Evaluation:
  • Accuracy
  • Precision, Recall, F1-score
  • Confusion Matrix

🚀 Technical Requirements

• Languages: Python
• Tools: VS Code / Google Colab / Jupyter Notebook
• Libraries: NumPy, Pandas, Matplotlib, Seaborn, Scikit-learn, XGBoost

📈 Key Insights

• The dataset had a balanced gender representation (~53% female, ~47% male).
• Labour had a higher voter count (69.7%) compared to Conservative (30.3%).
• Logistic Regression and XGBoost performed well, with accuracy around 85%.

📑 Deliverables

• Jupyter Notebook with full implementation.
• Business report in PDF format (excluding code).
• Visualizations and insights from model predictions.

🔗 References

• Project report: Report.pdf
• Dataset: ElectionData.xlsx
• Sample Notebook: SOT23B1_Abhyudaya.ipynb

---

Election_Voting_Prediction