PROJECTS

Server Load Prediction Model

Implemented Decision tree and Bagging classifier on server load data set. The bagging classifier with an F-score of 0.91 performed relatively better than the decision tree classifier with an F-score of 0.84.

Wisconsin Breast Cancer Classifier

Implemented Decision tree and Random forest classifier on Wisconsin breast cancer data set. Random forest with an F-score of 0.96 performed better than the Decision tree with an F-score of 0.86.

Cognitive Model of Perception of Emotions

• Simulated visual perception of emotions using deep Convolutional neural networks and Deep Dream algorithm. Found results consistent with the findings of psychological studies conducted on human beings.

• Used Keras API to create an emotion recognition model after training it on the FER 2013 data consisting of 7 classes of emotions. The F-score of the trained model was 0.6 on the test data set.

MOOC Certification rate vs Economic Status

• Used linear regression and visualization techniques to model MOOC data from MIT-X and Havard-X. Found correlation between certification per registration rate and course fee per GDP Capita with Pearson coefficient of – 0.52 and p-value of 0.009.

RASA Chat Bot

• Built RASA chatbot for Hindi language on WALS dataset using spacy library and embeddings from fasttext.cc.

• Fine-tuned GPT-2 language model on Tiny Shakespeare and Bill Sum corpora using HuggingFace set of tools. After fine tuning generated some text using the fine-tuned model.