Projects¶
Waste bin empty date prediction¶
Project was part of course work but customer was client
Model:¶
We experienced with two different time series models: - ARIMA (AutoRegressive Integrated Moving Average) - ETS (Error, Trend, Seasonality)
++ also calculated naive average model as benchmark
Libraries:¶
- python
- pandas
LASTFM telegram bot | github¶
Classifying cancer tissue types | notebook¶
Objective:¶
Our team explored different machine learning models to classify the primary tissue of origin in cancer samples by using mutational signatures— patterns of mutations associated with specific biological processes or exposures.
Tech Stack¶
Python Pandas NumPy Jupyter Notebook Scikit-learn Matplotlib, Seaborn, Plotly
My contribution:¶
Data preprocessing & visualization:¶
Cleaned and preprocessed the mutation signature dataset. Created visualizations to explore data distributions and relationships using Matplotlib, Seaborn, and Plotly.
Modeling:¶
Implemented the MLP and combined all of the other classification algorithms (KNN, SVM, Random Forest, SGD, Logistic Regression, Naive Bayes) into a cohesive pipeline. Conducted hyperparameter tuning to optimize model performance.
Project Management:¶
Suggested to use git for version control, ensuring reproducibility.
Key Results¶
Top Performer: The Multi-layer Perceptron (MLP) achieved a peak accuracy of 80% after hyperparameter tuning.
