Skip to content

[Data Science] This repository contains data science competitions ranging from topics like text classification to image recognition.

Notifications You must be signed in to change notification settings

k-chuang/data-science-competitions

Repository files navigation

Data Science Competitions

This repository contains data science competitions from a data science related graduate course. The competitions (2 supervised learning, 1 unsupervised learning, 1 recommender system) are separated into folders, and are discussed more in detail below.

Note: These competitions follow a similar format with kaggle competitions in terms of ranking & scoring. The public leaderboard ranking is evaluated on 50% of the test set, and the final leaderboard ranking is evaluated on the whole test set.

Description

Develop the K-nearest neighbors algorithm (supervised learning) from scratch to determine, given a medical abstract, which of 5 classes (medical conditions) it falls in. The 5 classes are diseases: digestive system diseases, cardiovascular diseases, neoplasms, nervous system diseases, and general pathological conditions

Dataset

Training dataset: ~14,400 raw medical text records with class labels (1-5) provided Test dataset: ~14,400 raw medical text records with no class labels

Rank & F1 Score

Public Leaderboard Rank: 2/45 with a F1 score of 0.7858

Final Leaderboard Rank: 5/45 with a F1 score of 0.7815

Applied Concepts

  • K-nearest neighbor algorithm
  • Text pre-preprocessing techniques (Bag of Words approach with tf-idf vectorization)
  • K-fold cross-validation

Description

Develop a predictive model (supervised learning) that can determine, given images of traffic depicting different objects, which class (out of 11 total classes) it belongs to. The object classes are: car, SUV, small truck, medium truck, large truck, pedestrian, bus, van, people, bicycle, and motorcycle.

Dataset

Training dataset: ~21,200 records of pre-extracted image features (HOG, Hist, LBP, RGB, DF features) in dense matrix format with class labels provided (1 - 11) Test dataset: ~5,300 records of pre-extracted image features in dense matrix format with no class labels provided

Rank & F1 Score

Public Leaderboard Rank: 1/45 with a F1 score of 0.8598

Final Leaderboard Rank: 2/45 with a F1 score of 0.8508

Applied Concepts

  • Ensemble methods (bagging, boosting, soft voting)
  • Feature selection techniques (variance threshold)
  • Dimensionality reduction techniques (PCA, Truncated SVD)
  • K-fold cross-validation
  • Bayesian optimization for model parameter tuning

Description

Develop the DBSCAN clustering algorithm (unsupervised learning) from scratch to cluster news articles (text).

Dataset

Training dataset: ~8,600 news articles in document-term sparse matrix format (word index, word frequency)

Rank & NMI Score

Public Leaderboard Rank: 2/45 with a NMI score of 0.6238

Final Leaderboard Rank: 3/45 with a NMI score of 0.6249

Applied Concepts

  • DBSCAN clustering algorithm
  • Dimensionality reduction techniques (PCA, Truncated SVD)
  • Measuring cluster validity with internal measures (Silhouette score, Calinski and Harabaz score, Cohesion with SSE)

Description

Develop a recommender system for books based on user ratings on books

Dataset

Training dataset: 700,000 ratings (user_id, book_id, rating)

Test dataset: 300,000 ratings

Rank & RMSE score

Public Leaderboard Rank: 1/40 with a RMSE score of 1.50603

Final Leaderboard Rank: 1/40 with a RMSE score of 1.51056

Applied Concepts

  • Recommender Systems
  • Collaborative Filtering
    • Neighborhood-based (or memory-based)
    • Matrix factorization (or model-based)

About

[Data Science] This repository contains data science competitions ranging from topics like text classification to image recognition.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages