https://github.com/asrot0/imdb_sentiment_analysis

https://github.com/asrot0/imdb_sentiment_analysis

Last synced: 2 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/asrot0/imdb_sentiment_analysis
• Owner: asRot0
• Created: 2025-02-17T15:14:43.000Z (2 months ago)
• Default Branch: main
• Last Pushed: 2025-02-17T15:47:55.000Z (2 months ago)
• Last Synced: 2025-02-17T16:28:22.282Z (2 months ago)
• Size: 3.91 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# IMDB Sentiment Analysis

## Project Overview

This project performs sentiment analysis on **IMDB movie reviews** using **K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Random Forest, and XGBoost**. The goal is to classify reviews as **positive or negative** based on textual content.  

### **Steps Involved**

1. **Load & Explore the Dataset**  

2. **Preprocess Text Data** (Cleaning, Tokenization, Stopword Removal, Stemming)  

3. **Train-Test Split** (70% Train, 30% Test)  

4. **Feature Extraction** using **Bag of Words (BoW)**  

5. **Train KNN Model & Evaluate**  

6. **Train SVM, Random Forest, XGBoost on a subset (40%-50%) & Compare**  

7. **Hyperparameter Tuning using RandomizedSearchCV**  

8. **Train the Best Model on the Full Dataset**  

9. **Evaluate the Final Model (Accuracy, F1 Score, Confusion Matrix, Visualizations)**  

---

## Dataset Details

- **Dataset Source**: [IMDB Dataset](https://github.com/asRot0/machine-learning/blob/main/datasets/IMDB%20Dataset.csv)  

- **Size**: 50,000 reviews  

- **Classes**:  

  - **Positive (25,000 reviews)**  

  - **Negative (25,000 reviews)**  

Each review is labeled as **positive** or **negative**, making it a **binary classification problem**.

---

## Technologies Used

- **Python**  

- **Pandas, NumPy** (Data Handling)  

- **Scikit-Learn** (Machine Learning Models)  

- **XGBoost** (Boosting Algorithm)  

- **Seaborn, Matplotlib** (Data Visualization)  

- **NLTK, BeautifulSoup** (Text Processing)  

---

## Model Evaluation & Insights

To understand model effectiveness, we analyzed **confusion matrices** and **classification reports** for each model. Below are some key insights:

### 🔹 **K-Nearest Neighbors (KNN)**

- Performed **poorly** due to the high-dimensional sparse nature of text data.

- Struggled with decision boundaries, leading to **low accuracy**.

### 🔹 **Support Vector Machine (SVM)**

- Provided **decent performance** with good generalization.

- However, training time was relatively **slow** on a large dataset.

### 🔹 **Random Forest**

- Showed **strong results**, handling non-linear relationships well.

- Benefited from **ensemble learning**, but had slightly **higher training time**.

### 🔹 **XGBoost**

- Achieved the **best accuracy**, excelling in feature selection & boosting weak learners.

- Benefited significantly from **hyperparameter tuning**.

- Final model trained on **full dataset** after parameter optimization.

### **Visualizing Model Results**

Here’s a heatmap of the **best model’s confusion matrix**:

```python

import seaborn as sns

import matplotlib.pyplot as plt

from sklearn.metrics import confusion_matrix

cm = confusion_matrix(y_test, y_pred_final)

sns.heatmap(cm, annot=True, fmt='g', cmap="Blues")

plt.title("Best Model (XGBoost) - Confusion Matrix")

plt.xlabel("Predicted")

plt.ylabel("Actual")

plt.show()

```

> **Note**: Hyperparameter tuning was performed on the best-performing model before final training.

---

## How to Run

### **1. Install Dependencies**

```bash

pip install pandas numpy scikit-learn xgboost seaborn matplotlib nltk beautifulsoup4 tqdm imbalanced-learn