🧠 Machine Learning Mini Projects

This repository contains a collection of machine learning mini projects developed during my academic studies and self-learning journey. Each project focuses on applying data preprocessing, model training, and result interpretation on real-world datasets.

---

📁 Projects Included

Flowers

• Description: Classifying 10 flower species based on image data using deep learning techniques.
• Dataset: Flower Classification | 10 Classes |
• Techniques used: Image preprocessing, data augmentation, transfer learning (e.g. VGG16, ResNet50, InceptionResNetV2), model evaluation with accuracy and confusion matrix.
• Goal: Compare performance of classification models.
• Result: ResNet50 achieved the highest accuracy at 74.45%.

AirPollution_Time_Series

• Description: Time series analysis and forecasting of air pollution data.
• Dataset: Air Pollution Forecasting - LSTM Multivariate
• Techniques used: Time series preprocessing, linear regression, LSTM, and model evaluation with MAE and RMSE.
• Goal: Compare performance of models on air pollution data.
• Result: LSTM had better performance (MAE: 90.22, RMSE: 102.61) than Linear Regression on air pollution forecasting.

VideoGameSales

• Description: This dataset contains a list of video games with sales greater than 100,000 copies. It was generated by a scrape of vgchartz.com.
• Dataset: Video Game Sales
• Techniques used: feature scaling (MinMaxScaler, MaxAbsScaler), regression modeling (Linear Regression, Random Forest, SVR), pipeline construction, hyperparameter tuning (GridSearchCV), and model evaluation with R².
• Goal: Compare performance of models, perform exploratory analysis to identify sales trends by region and platform.
• Result:
  • Best model: MinMaxScaler + Linear Regression (R² = 0.99998)
  • Region with highest total sales: North America (NA)
  • Platform with highest global sales: PlayStation 2 (PS2)

---

🛠️ Tools & Libraries

• Python, Pandas, NumPy, Matplotlib, Seaborn, Plotly
• Scikit-learn (Linear Regression, Random Forest, SVR, Pipeline, GridSearchCV)
• TensorFlow, Keras (CNN, LSTM, Transfer Learning)
• Jupyter Notebook, Google Colab

---

⭐ This project was developed during my learning journey and reflects my ability to apply concepts into practice. It continues to be improved as I grow