Arnab Mukherjee

Experienced telecommunication engineer with additional expertise in data science and statistics, successfully leveraging Python and machine learning to drive innovative solutions at Intellipaat Software Solutions. Proven ability to work under minimum support in fast-paced environments requiring attention to detail and leadership skills at NOKIA, strong analytical problem-solving skills, and data analytics to transform complex data into actionable insights at Bharti Airtel.

• MITx edX 6.86x: Machine Learning with Python https://courses.edx.org/certificates/7ba00d3f3210422ab00d42ef3c07ecf3
• TensorFlow 2.0: Deep Learning and Artificial Intelligence https://www.udemy.com/certificate/UC-e2ac04c4-e579-43ac-8087-37f522cc6f00/

• Diabetes Disease Detection:: Classified diabetes patients using features like glucose level, BMI, age, insulin, and blood pressure. Conducted Exploratory Data Analysis (EDA) with Matplotlib and Seaborn. Implemented Logistic Regression, Decision Tree,KNN, Random Forest and ANN using TensorFlow2. Tuning hyperparameters and best model selection by evaluating ROC-AUC curve. Best performing model is KNN with accuracy of 81%.
• Smoking Prediction:: Classified smokers vs. non-smokers using biological parameters from 55k individuals. Conducted data cleaning using pandas, EDA and visualizations with Pandas, Matplotlib, and Seaborn. Evaluated classification models (Logistic Regression, Decision Tree, Random Forest, KNN, SVM). Achieved 84% accuracy with Random Forest and 78% with ANN.
• Handwritten Digits Detection:: Used the MNIST dataset of grayscale images of digits (0 to 9). Applied Principal Component Analysis (PCA) for dimensionality reduction. Implemented Decision Tree (85% accuracy) and Random Forest (96% accuracy). Achieved 97.5% accuracy using Artificial Neural Networks (ANN) and 95% with CNN.
• Flight Price Analysis using Regression:: Analyzed 300k flight price records data based on stops, class, source/destination, time etc. Identified key trends: direct flights are cheaper, business class has a major price gap, and last-minute bookings cost more. Visualized insights using Matplotlib and Seaborn. Applied Linear Regression with L1 & L2 regularization, Decision Tree Regressor, and TensorFlow2 deep learning models.
• Review Ratings Predictor:: Performed natural language processing task using NLTK library to tokenize words, used word embedding after removing 'stopwords' and applying 'lemmatization' to convert words into their root form. Afterwards 'Bag of Word' and 'TF-IDF' technique were used to convert words into meaningful vectors. Lastly, Multiclass Classification models were used on the top of word embedded vectors to properly classify user comments into meaningful representation in ratings numbers from 0-5.
• Recommendation System model:: By using Amazon product dataset ratings of unrated items were predicted using Singular Value Decomposition(SVD) algorithm from surprise package, user-hidden feature and item-hidden feature interaction were captured taking 15 of such hidden features into account.

• Indian Classical Music
• Stargazing