A Portfolio of Python Projects
These will be my forays into learning Python for data science.
Each project will be tracked here with a quick note about what the project does.
Below are some highlights of each project:
Image-Based Fashion Recommender
An image-based style recommender on top of a clothing classification system (95% accuracy) employing Convolutional Neural Network and Nearest Neighbors leveraging AWS and Keras with TensorFlow as backend. A demo of the recommender can be found under "Flask_App". The recommender can be used as a more effective search engine. It can also be applied to any domain that needs image content detection, tagging or image-based recommendations.
Personal Email Management System
Personalized auto email grouping system applying NLP and unsupervised ML, with a focus on clustering and dimensionality reduction (topic modeling), including PCA, NMF, LDA, t-SNE, K-Means and Agglomerative Clustering. The system aims at clustering emails into different topics based on email content to help users work more efficiently.
Ingredient Based Cuisine Classification
Applying supervised ML algorithms, including Random Forest, Naive Bayes, SVM and Logistic Regression, to classify cuisines from 20 countries based ingredients on recipes.
Domestic Movie Gross Prediction
Built three predictive models for 2D, 3D, and all movies applying Linear Regression with Lasso/ Ridge/ ElasticNet Regularization, each yielded an adjusted R-square of over 0.95, using data scrapped from the BoxOfficeMojo.com leveraging BeautifulSoup.