# SBU Artificial Intelligence Final Project 2022 **Multiclass Text Classification | Natural Language Processing | Kaggle** This repository is dedicated to the [SBU Artificial Intelligence Final Project](https://www.kaggle.com/competitions/sbu-ai-finalproject/overview), a Kaggle competition focused on **multiclass text classification**. The challenge is to build a machine learning solution that assigns each input text to the appropriate category. Beyond producing predictions, this project emphasizes a systematic NLP workflow: exploring the data, preparing text, developing models, validating their performance, and generating competition-ready submissions. ## Objectives - Understand the dataset and explore the distribution of text samples and labels. - Prepare text data for machine learning through appropriate cleaning and preprocessing. - Develop and compare approaches to representing and classifying text. - Evaluate models on held-out data and reduce the risk of overfitting or data leakage. - Generate predictions in the format required by the competition. ## Workflow 1. **Data exploration:** Inspect the training and test sets, identify missing values, and examine class distributions. 2. **Text preprocessing:** Apply suitable normalization, cleaning, and tokenization techniques. 3. **Feature representation:** Transform text into features that can be used by a classification model. 4. **Model development:** Train and refine candidate models for multiclass prediction. 5. **Validation:** Compare experiments using an appropriate validation strategy and evaluation metrics. 6. **Submission:** Run inference on the test set and prepare a submission file according to Kaggle's specifications. ## Dataset The dataset is provided through the official competition page: **[View the competition data on Kaggle](https://www.kaggle.com/competitions/sbu-ai-finalproject/data)** Access may require signing in to Kaggle and accepting the competition rules. Refer to the dataset documentation for the available files, column names, and target labels. ## Getting Started 1. Clone this repository to your local machine. 2. Download the competition data from Kaggle and place it in your preferred local data directory. 3. Install the dependencies required by the project's notebooks or scripts. 4. Run the preprocessing, training, and prediction steps provided in the repository. 5. Verify that the generated submission matches the competition's required format before uploading it. If you use the Kaggle command-line interface, you can download the competition files with: ```bash kaggle competitions download -c sbu-ai-finalproject -p data ``` Make sure your Kaggle API credentials are configured and you have access to the competition. ## Evaluation and Submission Model performance should be assessed using a validation split that reflects the competition task. The **official scoring metric and submission format** are defined on Kaggle and should be treated as the source of truth. - [Evaluation details](https://www.kaggle.com/competitions/sbu-ai-finalproject/overview/evaluation) - [Competition leaderboard](https://www.kaggle.com/competitions/sbu-ai-finalproject/leaderboard) ## Acknowledgments This project was developed in the context of the **SBU Artificial Intelligence Final Project**. Thanks to the competition organizers and Kaggle for providing an opportunity to apply machine learning and natural language processing techniques to a practical classification task. --- *From raw text to meaningful predictions — one experiment at a time.*