rgpatel Goto Github PK

boston_housing_project

Boston housing data (from UCI ML Repository) is used to predict the prices of houses using Decision Tree Classifier. Performed data exploration (feature observation, identifying performance metric), training and testing of model (learning curves, complexity curves) and optimization of model (grid search, k-fold cross validation).

callysto-sample-notebooks

customer_segments_unsupervised_learning

Performed clustering using Gaussian Mixture Models to identify probable customer segments and suggested optimized A/B tests based on clustered data.

dog_breed_classifier_app

Used the dataset of 8300 dog images to identify the dog breed from 133 total breeds using convolutional neural network.

finding_donors_supervised_learning

Created a training and predicting pipeline based on AdaBoost algorithm to classify potential donors from US census data (45,000 entries) with 85% quality.

fly_quadcopter_rl

Designed a reinforcement learning agent to take-off quadcopter using DDPG – an actor critic method. Also, enhanced learning and repeatability of results by tuning reward function and hyper parameters.

google_landmark_challenge

Used dataset containing 21,000+ images of 200 unique landmarks around the world (from Google Landmark Recognition Challenge 2018) and developed convolutional neural networks using transfer learning providing up to 96% accuracy.

hackathon

Hackathon Challenges and Resources

imdb_movie_reviews

Predicted positive/negative sentiment of more than 25,000 movie reviews using “Bag of Words” processing with deep neural networks and achieved 80% accuracy.

publish-plotly-website

ynpreet

zeek-sniffpass

Sniffpass will alert on cleartext passwords discovered in HTTP POST requests