This repository is a smaller version of 100DaysofML to motivate beginners to take up that challenge and dive deeper into the domain of Machine Learning.
- Worked with StratifiedKFold and XGBoost for a private anomaly detection dataset. F1 - Score was found to be around 96%.
- Worked on Data Visualisations using Pairplot and Distplot to test Hypothesis and measure Skewness. Also worked on increasing Correalation of X with Y.
- Worked on Feature extractions for decreasing time consumed to train the model. Models used
- Worked with XGBoost on the dataset and finetuned to reach an F1-score of 96.2% on an anomalies detection dataset.
- Worked on reducing skewness of the dataset. Also worked with Power Tranform and understood the working.
- Right skewed: log transform
- Left skewed : square transform
- Learned EDA technique and and new types of graphs in Sea Born like swarmplot, KDE etc.
- Half way through writing research paper, learned about ResNet in depth, learned BoxCox for reducing skewness.
- Tried a skewness reduction method on Pet Adoption data. Learned more about hypothesis and P-value.
- After multiple tests on Training set, I applied the mehtods to test datatset. Sadly the accuracy achieved was only 88%. Worked more on hypothesis understanding.
- Researched about autocorrect methods using Deep Learning and distance method.
- Trying to implement Bi Directional LSTM
- Back to basics for Multilayer perceptron from scratch. Associative Network trial in Matlab
- Recieved Assignment related to geospatial data for internship. Data Analysis + Classification
- Read and understood these research paper
- Read a number of blogs: