Data Science portfolio

Robotic Arm Project

ROS is a Robotic Framework used for devoloping Robots. The video below shows how you can use ROS framework to with RViz to control a Robotic Arm. To achieve this, we first need to code up the plan for the robotic arm movement. This enables us to control the different joints of the arm and ensure the trajectory planning is stable.

https://github.com/Dom88Finch/DS_portfolio/assets/57221218/218b6d97-337e-4688-be51-f55d6a627f3f

In-play – Tennis Prediction project

Tennis is a popular sport that’s played across the world. The web application below shows an example of how we can predict in-play match outcomes by using Ensemble Machine Learning models. The apps enables for prediction of set winner or game winner by leveraging historical event data of major tennis tornaments such as Wimbledon. App is build using python libraries such as streamlit, Data Science Stack, HTML, CSS.

https://github.com/Dom88Finch/DS_portfolio/assets/57221218/d7b133b0-954d-4e4e-9ff4-1e588cbd0205

DL Audio: Vowel & Gender classification using Multi-Task Deep Learning Architecture

Multi-Task Convolutional Neural Network used to classify audio for vowel and gender. The repository contains a detailed analysis report of several different neural network architectures.

##

Sales Analysis using Excel and Power BI

Sales data including visualization with Power BI

• Obtain sales data from company online.
• Carry out statistical analysis of data and identify possible trends or area to divert resources.
• Data modeling using DAX functions.
• Created Visuals that enable the user to be able to answer key business questions

screenshot showing dashboad analysis of profit

Project X :Alpha signal generation for equity & crypto-currency markets using AI

*Aim: To be able to produce entry and exit signals for multiple assets across multiple markets with an expected return of 6-9% weekly return. This is then implemented across a fully automated risk management system that is dynamically adjusted using reinforcement learning algrorithms. *

Learning objectives and software

• Finance basics - Fundamental analysis, Technical Analysis
• Best investments strategies (short and long term)
• Trader Psychology
• Systematic trading
• Risk management systems

## other projects

Project 1:Data Science Student Grade predictor: Project Overview

• Created a mini app that predicts the student’s grade for their Final year exam.
• Used linear regression to achieve the best model achieving a 90% accuracy score
• Built a GUI for users to interact with using tinker

Project 2: Hospital stay prediction

Predicting duration of stay in hospital during this covid-19 pandemic to help with planning and ensuring there is enough hospital capacity to accommodate the patients.

• Used feature selection techniques to identify best feasutes to include in model
• Optimized Logistic, Random Forest using GridSearchCV to imporve model
• model prediction score of 0.35

Project 3: Dating website analysis

Analysis of OKCupid dataset to find possible relationships and indicators that may help with matching individuals together.

• invlolved data cleaning and preprocessing data
• Engineered features form the text from multiple columns to improve comparison analysis between individuals
• Use NLP techniques to analyse the description column of the data

Currently researching what make coupples get attracted to one another and also working on a scoring matrix that matches individuals based on a a number of features including:

• Age, Likes and dislikes, Ethnicity, Similar Habits, Personality traits

Project 4: Classifying tweets

Program that uses Machine learning to classify whether a tweet is viral or not. The program initially uses a KClassifier to classify and predict whether a tweet is viral. The program is then modified preprocessing the data frame again to find new potential feature data that would improve our prediction.

This involved cleaning the data using python packages. Choosing a minimum score with which to classify whether a tweet is viral or not. Chose the median to be the viral score minimum of 13 for a tweet to be classed as viral. Created new variables which hold Boolean values to act as training set for the machine learning model. I then trained my model to predict whether a tweet was viral. From this I did further investigation to improve my model by choosing different training feature data to improve the accuracy. I was able to improve it by 15%. )