The problem Machine Learning Project Portfolio solves

This portfolio showcases the application of machine learning techniques to solve diverse problems:

New BRICS Member Sentiment Analysis: Predicts the sentiment of text regarding the inclusion of new members in the BRICS alliance, aiding policymakers in understanding public opinion.
AirBus Ship Detection: Detects ships in satellite images, enhancing maritime surveillance and security.
Customer Review Sentiment Analysis: Analyzes sentiments in customer reviews, helping businesses understand customer feedback and improve products/services.
Mineral Classification: Classifies different types of minerals based on their properties, facilitating geological studies and resource management.
Used Car Price Prediction: Predicts used car prices based on various attributes, assisting buyers and sellers in making informed decisions.
Health Insurance Cross-Sell Prediction: Predicts whether a customer will buy a health insurance policy, helping insurance companies target potential buyers effectively.

Challenges I ran into

Each project posed unique challenges:

New BRICS Member Sentiment Analysis: Major challenge was converting text data into a format suitable for model fitting. Extensive text preprocessing like tokenization and stemming was required.
AirBus Ship Detection: Handling large satellite images and ensuring accurate segmentation of ships were significant hurdles.
Customer Review Sentiment Analysis: Dealing with noisy and unstructured text data, and balancing the dataset for accurate sentiment classification was challenging.
Mineral Classification: Ensuring model generalization with a small dataset required careful handling of features and cross-validation.
Used Car Price Prediction: Managing diverse attributes and handling missing values while building a robust regression model posed challenges.
Health Insurance Cross-Sell Prediction: Imbalanced classes in the dataset and choosing the right evaluation metrics were key issues to address.