This repository contains the tasks I completed during my 2-week remote internship on Machine Learning.
The internship required a minimum of 3 tasks, but I successfully completed all 8 tasks to gain broader practical experience.
Each folder corresponds to one task and contains the code, datasets (if applicable), and results.
├── Task1_StudentScorePrediction/
├── Task2_CustomerSegmentation/
├── Task3_ForestCoverClassification/
├── Task4_LoanApprovalPrediction/
├── Task5_MovieRecommendation/
├── Task6_MusicGenreClassification/
├── Task7_SalesForecasting/
└── Task8_TrafficSignRecognition/
- Built a model to predict students' exam scores based on study hours and related factors.
- Performed data cleaning, visualization, and train-test split.
- Trained a Linear Regression model, visualized predictions, and evaluated performance.
- Bonus: Explored polynomial regression and experimented with feature combinations.
- Clustered customers into segments based on income and spending score using the Mall Customer dataset.
- Applied K-Means clustering, determined the optimal number of clusters, and visualized results.
- Performed scaling and exploratory analysis.
- Bonus: Tested alternative algorithms like DBSCAN and analyzed spending patterns per cluster.
- Predicted types of forest cover using environmental and cartographic features from the UCI Covertype dataset.
- Cleaned and preprocessed categorical data.
- Trained and evaluated multi-class classification models.
- Visualized confusion matrices and feature importance.
- Bonus: Compared models such as Random Forest and XGBoost with hyperparameter tuning.
- Built a model to predict loan application approvals.
- Preprocessed data: handled missing values and encoded categorical features.
- Focused on imbalanced data challenges, using metrics such as precision, recall, and F1-score.
- Bonus: Implemented SMOTE for balancing and compared logistic regression vs. decision tree models.
- Developed a user-based collaborative filtering system using the MovieLens dataset.
- Constructed a user-item matrix to compute similarity scores.
- Recommended unseen, top-rated movies for users.
- Evaluated recommendation quality using Precision@K.
- Bonus: Experimented with item-based filtering and matrix factorization (SVD).
- Classified songs into genres using the GTZAN dataset.
- Extracted features (e.g., MFCCs) with Librosa and tested both tabular and image-based approaches.
- Implemented multi-class models using Scikit-learn and CNNs for spectrogram images.
- Bonus: Applied transfer learning for CNN models.
- Forecasted future sales using Walmart’s historical sales data.
- Engineered time-based features such as lags, rolling averages, and seasonal decomposition.
- Trained regression models and visualized predicted vs. actual sales over time.
- Bonus: Tested XGBoost/LightGBM with time-aware validation.
- Built a CNN model for traffic sign classification using the GTSRB dataset.
- Preprocessed images (resizing, normalization).
- Trained and evaluated deep learning models with accuracy and confusion matrices.
- Bonus: Compared performance of a custom CNN vs. pre-trained models (e.g., MobileNet) and used data augmentation.
- Python
- NumPy, Pandas, Matplotlib, Seaborn
- Scikit-learn, XGBoost, LightGBM
- Librosa (for audio features)
- TensorFlow/Keras (for CNN models)
- OpenCV (for image preprocessing)
- Completed all 8 tasks in just 2 weeks.
- Worked with real-world datasets (student performance, mall customers, UCI Covertype, MovieLens, GTZAN, Walmart sales, GTSRB).
- Covered a wide range of ML topics: regression, classification, clustering, recommendation, forecasting, and deep learning.
-
Clone the repository:
git clone https://github.com/username/repo-name.git
-
Navigate into any task folder.
-
Upload or provide access to the dataset:
- Option A: Upload directly to Colab during runtime
from google.colab import files uploaded = files.upload() - Option B: Mount Google Drive and access the dataset from there
from google.colab import drive drive.mount('/content/drive')
- Option A: Upload directly to Colab during runtime
-
Open the Jupyter Notebook or Python script to explore the code and results.
Thanks to Elevvo for the opportunity to apply and expand my skills through this internship.