Navkaran Singh Portfolio

Scraping a classic novel, extracting its text, and analyzing word distribution using NLTK to unlock insights from unstructured data.

• View Project

Discovered trends in app categories, pricing strategies, and user sentiment. Dive deep into the data to make informed decisions about app development and marketing.

• View Project

Analyzing player retention in "Cookie Cats" to determine the optimal gate placement for maximizing engagement and retention rates.

• View Project

Transforming Netflix Movie Data into Insights: A Comprehensive Exploratory Data Analysis (EDA) Project Focused on the 1990s. Uncover Key Patterns, Trends, and Insights in the Film Industry Using Python, Pandas, Matplotlib, and Seaborn.

• View Project

Uncovering insights into the performance of a bike-sharing company through data integration, SQL queries, and advanced data visualization techniques. This project showcases a comprehensive Power BI dashboard to facilitate actionable business decisions.

• View Project

Dive into the world of recommendation systems with this comprehensive project on Alternating Least Squares (ALS) in PySpark. Using the MovieLens and Million Songs datasets, this project demonstrates data preprocessing, model training, hyperparameter tuning, and model evaluation, showcasing the power of ALS for generating meaningful recommendations.

• View Project

Developed a model to predict captions for images using the Flickr8k dataset, employing CNN for image processing and LSTM for text generation. Achieved a BLEU-1 Score of 0.544 and a BLEU-2 Score of 0.319.

• View Project

Conducted a comprehensive time series analysis and forecasting for furniture sales, utilizing ARIMA for forecasting and Prophet for modeling complex patterns. This project involved preprocessing data, visualizing time series, validating forecasts, and comparing trends.

• View Project

Developed a machine learning model to classify songs into Rock or Hip-Hop genres using track information from Echonest/Spotify. The project includes data preprocessing, dimensionality reduction with PCA, and classification with Decision Trees and Logistic Regression. The trained models are saved as .pkl files and used with librosa for genre prediction from MP3 files.

• View Project

Developed a reinforcement learning Q-learning agent to optimize urban transportation in the Taxi-v3 environment by training over 2,000 episodes. Implemented an ε-greedy strategy and fine-tuned hyperparameters to enhance decision-making in a 5x5 grid world. Analyzed Q-values and policy, with future exploration into multi-agent scenarios and real-world autonomous navigation application.

• View Project