This page showcases a variety of my projects in Data Science, AI, NLP, and computer vision. You’ll also find Kaggle datasets that I’ve curated and preprocessed for my research. To see more of my work, visit my GitHub page, download my CV, or check out the About page.
My portfolio features the following projects:
🏡 Airbnb NYC Price Prediction (2024).
📰 Fake News Detector using NLP & Machine Learning.
🧠 MCI to AD MRI Classification.
🎭 Sentiment Analysis on IMDb Reviews.
Airbnb NYC Price Prediction (2024)
Project Highlights :
Built a machine learning model to predict Airbnb prices in New York City using data cleaning, feature engineering, and advanced regression techniques.
Compared two models, Linear Regression, Random Forest.
Provided data-driven insights on how location, room type, and host activity influence Airbnb pricing, helping optimize pricing strategies.
Tags: machine learning, regression models, data visualization, price prediction, exploratory data analysis
Fake News Detector using NLP & Machine Learning
Project Highlights :
Developed a machine learning pipeline to classify news articles as real or fake using NLP techniques like TF‑IDF vectorization.
Trained and evaluated multiple models (Naive Bayes, Logistic Regression) to achieve high classification accuracy.
Tags: natural language processing, machine learning, web app, text classification, Flask, scikit‑learn
Mild Cognitive Impairment to Alzheimer's Disease MRI Classification
Project Highlights :
Implemented a transfer-learning pipeline using EfficientNetB0 (pre‑trained on ImageNet) to classify MRI scans, distinguishing Mild Cognitive Impairment (MCI) from Alzheimer’s Disease (AD) with significantly improved feature extraction.
Packaged the model into a usable tool: incorporated comprehensive MRI preprocessing, fine-tuning, and evaluation (accuracy, precision, recall, ROC-AUC), positioning it as a strong candidate for clinical-support applications.
Tags: deep learning, medical imaging, transfer learning, EfficientNetB0,TensorFlow.
Sentiment Analysis on IMDb Reviews
Project Highlights :
Built an end-to-end NLP pipeline: Preprocessed 50,000 IMDb reviews (cleaning, tokenization, removing noise), converted text to numerical features, and engineered features like review length and sentiment scores.
Evaluated multiple classifiers: Trained models, including Naive Bayes, with performers achieving 88% accuracy, showing effectiveness across model types
Tags: natural language processing, sentiment analysis, text classification, IMDb dataset, machine learning, Flask, scikit‑learn.
My Kaggle datasets:
🧬 HPA Annotated Merged Dataset.
🩻 MCI to AD MRI dataset.
HPA Annotated Merged Dataset
Dataset Highlights :
Source & Purpose: The dataset comes from the Human Protein Atlas (HPA) project in Sweden, which aims to map proteins in all human cells, tissues, and organs.
Content & Structure: It includes 89,236 images of human cells, each represented by four color channels (green for proteins, blue for the nucleus, red for microtubules, and yellow for the endoplasmic reticulum).
Annotations: A train.csv file with 21,806 records links image IDs to one or multiple labels, representing different subcellular structure classes.
MCI to AD MRI dataset
Dataset Highlights :
Source & Content: The dataset is provided by the Alzheimer’s Disease Neuroimaging Initiative (ADNI), funded by the NIH. It includes two heterogeneous data types: MRI scans (97,036 images) and gene expression profiles (49,395 records), collected from 811 participants across multiple cognitive states (AD, MCI, CN, etc.).
MRI & Genetic Data: MRI data were collected across three ADNI phases (ADNI1, ADNI2, ADNI3), using evolving protocols (e.g., T1-weighted scans, 3T scanners). Genetic data were obtained using Bead-Studio 3.2 and Illumina Bead Chips with over 730,000 SNP markers.
