Portfolio

A short description to my projects related to Data Analysis and Machine Learning.

Machine Learning Competitions (Kaggle / Zindi)

To Vaccinate or Not to Vaccinate: In this hackathon, a 3-class model is needed to determine whether the text is pro-vaccination, anti-vaccination or neutral.
- Trained model: https://drive.google.com/open?id=17NGDdUfn2nEEkgudsnndyFgNm3MZDJX8
- Position: 7 out of 130 contestants
- Contest: URL
- Tools: Jupyter Notebooks - Python 3 - BERT - transformers - Google Colab
Financial Inclusion in Africa: In this binary classification task, One should implement a model to predict whether a person has a bank account or not given some information about the person.
- Position: 120 out of 643 contestants
- Contest: URL
- Tools: Jupyter Notebooks - Python 3 - xgboost - pandas
AWS DeepRacer Challenge: DeepRacer is a 1/18 scale car trained using Reinforcement learning to navigate a race track using footage from a camera. AWS provides a way to train the RL model by providing a reward function that determines the reward for a given state of the car. The state of the car is determined by multiple parameters such as: speed, distance to the centerline of the track, current position of the car(x, y).
- Position: 193 out of 1291 contestants
- The 3-lap demo for the model

Covid-19 Bio-NER: A notebook that fine-tunes BERT-based models to work on token classification task (Name Entity Recognition). Tools: Jupyter Notebooks - Python 3 - transformers

Analyze A/B Test Results: Investigate the results of an A/B test run by an e-commerce website to conclude whether to use the new design of the website or not.
This project is part of the Udacity's Data Analyst Nanodegree.
Tools: Jupyter Notebooks - Python 3 - pandas - numpy - statsmodels
Investigate the no-showup Dataset: Kaggle's no-showup dataset is explored in order to find if there is a dependancy between whether a patient shows-up in an appoinment and various attributes such as: patient age - period between Scheduling and appointment time.
Various plots are used to support the analysis process.
This project is part of the Udacity's Data Analyst Nanodegree.
Tools: Jupyter Notebooks - Python 3 - Matplotlib
Explore Bike Share Data: In this project, exploratory analysis is performed on data provided by Motivate, a bike-share system provider for many major cities in the United States.
The whole Data Analysis process is done from posing questions to drawing conclusions.
This project is part of the Udacity's Data Analyst Nanodegree.
Tools: Jupyter Notebooks - Python 3 - Matplotlib
Explore Weather Trends: Pull average Temperature of Cairo,Egypt and the worldwide average for the last 200 years and explore the changes and the general trends.
This project is part of the Udacity's Data Analyst Nanodegree.
Tools: Python 3 - SQL

Three Parity Problem: Use neural networks to find the parity of three bits and investigate the effect of increasing the no. of hidden nodes on the output.
This project is part of Neural Networks Undergraduate Course (CSE 463) taught at Faculty of Engineering - Ain Shams University.
Tools: GNU Octave
Radial Basis Function Clustering: Use backpropagation to optimize the parameters of the RBF
and investigate the effect of increasing the number of the means(RBFs) on the accuracy.
This project is part of Neural Networks Undergraduate Course (CSE 463) taught at Faculty of Engineering - Ain Shams University.
Tools: GNU Octave

Music or Speech: Develop a model for classifying audio files into either music or speech.
This model was developed as part of a 1-hour technical task for a deep learning position. Data couldn't be shared since I don't have the licence to use it. Tools: Python 3 - Scikit-learn - Seaborn
Maximum Aposteriori Classification: Apply MA classifcation to solve a simple 3D classification problem and visualize the decision boundaries.
This project is part of Selected Topics in System Engineering (Machine Learning) Undergraduate Course (CSE 465) taught at Faculty of Engineering - Ain Shams University.
Tools: Python 3 - Matplotlib
Detect Spam emails: Use Naive Bayes classification with laplacian smoothing to detect spam emails.
This project is part of Selected Topics in System Engineering (Machine Learning) Undergraduate Course (CSE 465) taught at Faculty of Engineering - Ain Shams University.
Tools: Python 3
SVM Classification: Explore the power of Support Vector Machines through a simple classifcation problem.
This project is part of Selected Topics in System Engineering (Machine Learning) Undergraduate Course (CSE 465) taught at Faculty of Engineering - Ain Shams University.
Tools: Python 3 - Scikit-learn