/HackerEarth-problems

Code for challenges/competitions hosted on HackerEarth

Primary LanguageJupyter NotebookMIT LicenseMIT

HackerEarth-problems

This repo contains code for challenges/competitions hosted on HackerEarth. These might include practice competitions and/or competitive challenges. Below each problem, I have listed different concepts/lessons which were learnt during the course of the challenge.

I. IndiaHacks 2017 Machine Learning - Predict the Road Sign : http://bit.ly/2iM1KAm

  • Basic structure and approach of an ML problem

II. ML Challenge #2 - Funding Kickstarter projects : http://bit.ly/2AU4DGP

  • Pre-processing of text features
  • CBOW + Naive Bayes/Tree models

III. Deep Learning Challenge #1 - Identifying grocery product category : http://bit.ly/2iHNlVR

  • Deep CNN models with Dropout, Data augmentation
  • Using ImageNet pre-trained models (VGG16 etc.)

IV. ML Challenge #4 - Predict Network Attacks : http://bit.ly/2BFx8Fp

  • Cross-validation setup
  • Feature engineering
  • Ensembles (stacking GBM models)

V. ML Challenge for Beginners - Predict Customer Happiness (Sentiment Analysis) : http://bit.ly/2kiRWON

  • Word cloud for EDA
  • Different text embeddings - word2vec, tf-idf, GLoVE etc.

VI. BrainWaves ML Challenge - Predict Annual Returns + Predict Fraudulent Transactions : http://bit.ly/2pvp6wL

  • Gradient boosting methods
  • Feature selection and Feature engineering (using xgbfir)
  • Stacked ensembles for regression