My Projects

Credit Card Fraud Detection

Fraud detection out of 284,807 transactions with high accuracy.

The dataset is highly unbalanced, the positive class (frauds) account for 0.172% of all transactions.

Dimensionality of the data is reduced for protecting the privacy of the users.

Implementing different machine learning models in a comparative fashion.

Building an Evaluation Matrix and different performance measure tools on Test Set

Face Detection

Creating a website that uses a machine learning model that classifies people's images.

This can classify images when images are uploaded to the website.

Data collecting and preprocessing is a very crucial step of the process.

Real Case Implementation of Convolutional Neural Networks.

Using differnet model evaluation tools and libraries and deployment of the model in a real dataset.

Real Estate Price Prediction

I have developed a machine learning model with a python flask server that predicts real estate prices according to their features.

Model allows user to enter home square ft area, bedrooms, etc and then retrieve the predicted price with very high accuracy.

This model can be adoptable to any kind of data that needs price predictions like car price, house price, item price, etc.

Different regression algorithms of machine leanring models is used in order to get the highest accuracy.

Customer Segmentation

This project uses unsupervised machine learning algorithms to perform customer market segmentation.

The project can enable companies to launch targeted ad marketing campaigns that are tailored to customer's specific needs.

Implemeting different segmentation strategies based on geographic, demographic, behavioral,psychological parameters.

The project can enable companies to develop optimization in budgeting, product design, promotion, marketing , and customer satisfaction.

Traffic Sign Detection

This project uses Convolutional Neural Networks (CNNs) to classify different traffic signs.

This model can classify 43 traffic sign images with a very accuracy and can be practically applied to self-driving cars.

Testing different filter sizes, depth, and number of convolution layers, as well as the dimensions of fully connected layers.

Using augmentation techniques in an attempt to extend dataset and provide additional pictures in different lighting settings and orientations.

Deploying Highly configurable code and developing a flexible way of evaluating multiple architectures.

My Personal Portfolio Website

I have developed and developed a personal and interactive portfolio website on my own that can be reached anywhere in the world.

This website is a fully fledged website that I developed by using both frontend and backend codes.

This website has a python flask server that serves different pages. The websites contains homepage, cv page, project page, reference page, and course page.

The website uses bootstrap and professional css Implementation.

In addition to data science skills, I can develope fully fledged websites by using HTML, css, javascript, flask, bootstrap, etc.

Customer Churn Prediction

I have developed a machine learning model by using different machine learning algorithms that can predict which customers are at high risk of churn with a very accuracy.

Model allows companies to detect both loyal and churn customers in a predictive way so that companies can develop a holistic view of the customers and their interactions across numerous channels.

This model can be adoptable to any kind of busines problem related to customer loyalty or churn.

Different classification algorithms of machine leanring models is used in order to get the highest accuracy.

Chatbot

This chatbot can conduct an online chat conversation via text or text-to-speech by using NLTK, Tensorflow and NLP.

The chatbot automates most of the customer interaction by answering some of the frequent questions that are asked by the customers.

The chatbot can be applicable to any kind of business domain.

I use a special recurrent neural network (LSTM) to classify which category the user’s message belongs to and then we will give a random response from the list of responses.

The chabot can be deployed to a website or an app.