StudyGroup

Developed a predictive model using machine learning to classify students into risk groups based on attendance, grades, and course difficulty. The model identifies failure probabilities, highlights key influencing factors, and includes a notification system for proactive management.

• Chevron Right

NHS

Analysed NHS appointment data alongside Twitter engagement to uncover patterns in healthcare delivery. Key findings include seasonal trends, high-traffic service settings, and the impact of wait times on attendance. Additionally, social media insights highlighted engagement gaps, emphasizing the need for more targeted strategies to enhance patient care and resource management.

• Chevron Right

Gaming Company

This project analysed customer loyalty points at a Gaming Company to uncover factors influencing customer behaviour. Using Python and R, I explored income, spending, and age data to develop predictive models. These insights were transformed into actionable strategies for improving engagement and optimizing marketing efforts.

• Chevron Right

• Chevron Left

In this study, I concentrated on identifying meaningful variables to estimate student performance effectively.By leveraging demographic data and course-related information, I developed classifications to better understand performance patterns. For instance, I used machine learning techniques like K-means clustering to classify course difficulty based on metrics such as pass rates and grade skewness.

Using the results, I developed a predictive model utilizing the random forest method. Here, you can see a comparison of the predicted outcomes versus the actual results on the test subset.Additionally, I created a probability calculator based on this predictive model, along with feature importance graphs for the first phase of the analysis.

In Phase 2 of the model, I incorporated attendance and grade performance to code the risk groups.After clustering students into these risk groups, I calculated the final feature importance map.The final product predicts failure probabilities and sends notifications to student managers when a student's status becomes concerning.

• Chevron Left

• Chevron Left

Through a detailed analysis of NHS appointment data, I uncovered distinct patterns in healthcare usage.Notably, November 2021 and March 2022 experienced significant peaks in appointments, highlighting periods of higher demand. General Practice emerged as the most frequented service setting, showcasing its critical role in patient interactions.Attendance rates remained strong overall but showed a slight decline when wait times between booking and appointments increased.

The data revealed that autumn and spring had the highest appointment counts, while summer showed surprising consistency in attendance rates. This stability suggests external factors, such as staffing or patient availability, may be influencing other months.Additionally, locations with consistent appointment durations demonstrated efficient scheduling practices, though the number of surgeries available didn’t necessarily translate to improved appointment stability.

Analysing Twitter data provided an additional lens into patient engagement. Most tweets showed low interaction, while those with broader announcements or promotions garnered higher engagement.Hashtags commonly reflected themes of health, technology, and medicine, yet actionable insights remained elusive.This highlights the untapped potential of social media as a tool for understanding patient sentiment and improving healthcare delivery strategies.

• Chevron Left

• Chevron Left

The analysis began with data cleaning and preparation using Python, followed by exploratory data analysis (EDA) in both Python and R.Word clouds and sentiment analysis highlighted themes in customer reviews, showing a strong positive sentiment around gaming experiences.Visualizations further revealed key trends, such as positive correlations between income, spending, and loyalty points.

Advanced statistical modelling was applied to predict loyalty points. A decision tree regressor in Python and multiple linear regression in R identified income, spending score, and age as significant predictors, with a model accuracy of 84%.Clustering techniques segmented customers into distinct profiles based on income and spending, providing valuable insights into customer behaviour.

The findings informed strategic recommendations, such as implementing tiered loyalty programs and targeted marketing campaigns for high-value customers.Personalized incentives and deeper segmentation analysis were proposed to boost engagement across customer segments.These insights can drive data-driven decision-making to enhance customer loyalty and business performance.

👨👩👧 That being said, I'm 40 years old, married to a wonderful woman, and a proud dad to a 4-year-old daughter who keeps life endlessly entertaining.

🏀🏐 I played basketball and volleyball competitively for over a decade as a licensed player. Staying active has always been a big part of my life, and while I don’t play as often now, those years on the court taught me a lot about teamwork and perseverance.

📷 For a while, I loved going on photography trips to capture unique perspectives of the world. While I haven’t been out with my camera recently, I still enjoy revisiting some of my favourite shots, which you can see here.

🎮 Gaming has always been a part of who I am. During my World of Warcraft days, I achieved a top-five warrior ranking and spent countless hours mastering FPS games like Counter-Strike.Beyond being a fun escape, gaming has also been a creative outlet. I used my analytical skills to develop custom World of Warcraft addons, one of which surpassed 30,000 downloads—a small but proud achievement that reflects my problem-solving mindset and love for building tools that others find useful.