-
π± Iβm currently learning Automation scripts using python
-
π I regularly write articles on Medium
-
π« How to reach me [email protected]
- Natural Language Processing
- Image Processing
- Visualizations
- Predicting real world outcomes
Classifying Sexual Harassment using Machine Learning
Following the #MeToo movement we had a lot of people opening up about their sexual harassment incidents, but as with any internet viral movement, it faded with time. Using the same data, worked on this case study and was able to achieve an average accuracy of 80% just two percent short of the original paper on the same topic.
Identifying emotions from tweets
Emotions are expressed through words, gestures, expressions, and with the ease of accessibility of social media today, emotions can now, also be expressed through tweets and Instagram/ WhatsApp stories. The project involved working on text cleaning, preprocessing, vectorization and modelling. I was able to achieve an accuracy of 82% on the test dataset with my approach.
Predicting the rating of a show based on it's plot and other metadata
Worked on this interesting intersection of NLP and regression problem to create a model that predicts the rating of a show based on it's written plot and other metadata. Worked on data analysis and modelling. Was able to achieve a public RMSE of 0.15750 with my code. Currently I am leading the public leaderboard with my score.
Classifying personalities using their images
As part of this article, worked on collecting data, and analysing it for image classification. Used openCV for feature generation and CNN for modelling, handled overfitting in the model using dropouts and image augmentation.
Visualizing Audio Data and Performing Feature Extraction
Audio is nothing but simply sound, the sound we hear in our daily lives. The human brain, through experience, can perfectly differentiate between the sound of an ambulance and the sound of a tabla, for example. But can we create machine learning models to work on such a classification? Working with audio data becomes a little overwhelming because we cannot visualize it as we can a set of tables or images. In this project, I worked on visualizing audio data followed by extracting useful features from the audio.
Football Match Probability Prediction
The dataset had nearly 164 features, as part of this project I worked on analysing these features to gain better insights into the data to ultimately create a model that would predict the chances of a football team winning a match!
ML Olympiad - Autism Prediction Challenge
As part of this project built a machine learning model to classify whether individuals have Autism or not. It was a public Kaggle competition, secured the first rank and amazing goodies from Google Developers Team for my work in this competition. My approach involved data analysis, feature selection and modelling with hyperparameter optimization:)
Predicting Customer Churn
Worked on an in depth analysis of the dataset, post which worked on modelling. Helped me gain a business understanding in the domain of customer churn.