AI ML Projects

This repository contains various AI and machine learning projects showcasing different techniques and approaches for solving real-world problems.

Current Projects

1. Butterfly Image Classification

• Task: Classifying butterfly species using deep learning.
• Approaches: Fully Connected Neural Networks (FCN), Convolutional Neural Networks (CNN), and Transfer Learning (ResNet152V2).
• Dataset: Butterfly Image Classification Dataset.
• Results:
  • FCN: Accuracy ~19.47%, Loss ~3.88
  • CNN: Accuracy ~85%, Loss ~58%
  • ResNet152V2: Accuracy ~91.50%, Loss ~34.93

---

2. Twitter Sentiment Analysis

• Task: Sentiment analysis on tweets to classify them as Positive or Negative.
• Model: RNN with Bidirectional LSTM.
• Dataset: Twitter Sentiment Analysis Dataset.
• Results:
  • Accuracy: 91.24%
  • Loss: 0.2137
  • Precision (Positive): 91%, Recall (Positive): 91%
  • Precision (Negative): 92%, Recall (Negative): 92%

---

3. Google Play Store User Reviews Analysis

• Task: Analyzing user reviews to classify sentiments (Positive, Negative, Neutral).
• Model: Logistic Regression using Spark ML Pipeline.
• Dataset: Google Play Store Apps Dataset.
• Results:
  • Balanced Precision, Recall, and F1-Score across Positive and Negative sentiments.
  • Sentiment distribution revealed a predominance of positive reviews.
  • Word clouds highlighted key terms:
    • Positive: "easy," "great," "love"
    • Negative: "poor," "crash," "bug"

---