Glucose Data Analysis

This project analyzes changes in glucose, insulin, exercise, and instantaneous glucose trends in time series data

---

Note: This project is a work in progress and will have new findings/investigations explained in the highlights section (below). Also, refer to screenshots of graphs below for an glimpse into the work done

---

Motivation

• To put my data analytics skills to the test, I wanted to work with real world data to find insights in the data that would help others
• The data used in this project comes from someone in my life that is important to me (I will keep them anonymous on their request)
• This person is a diabetic. I wanted to see if I could find insights in past data or create predictive models to help them deal with their diabetes better.

---

Tech/Framework

Data Analysis Done Using

• Python
• Pandas
• Scikit-learn (for data cleaning and models)
• Plotly (interactive graphs)
• Seaborn
• Matplotlib
• NumPy
• SciPy (statistical models and hypothesis testing)

Highlights

• Making functions to clean over 180 excel sheets containing glucose and exercise data
• Investigating daily summary statistics over the last 4 months by plotting average exercise and glucose on the same axes to visualize trends
• Exploring data for March specifically and finding statistically significant results between walking amounts and low blood glucose levels at night using levene and t-tests for hypothesis testing
• Researching the effects of instantaneous trend changes in glucose levels in March to see how instantaneous trends affect volatility of glucose over longer periods of time
• Identifying anomalies in less than 1% of the glucose time series data by using the interquartile range method, as well as Isolation Forest, Density-Based SCAN, and SVM models to understand the causes of past dangerous glucose levels

---

Notebooks/Files

1. cleaning_data.py: File that contains data transformations from excel sheets -> single dataframe
2. summary_exploration.ipynb: Notebook that contains the graphs of the summary information (data aggregated daily)
3. march_analysis/ipynb: Notebook that contains the walking and low blood glucose statistical hypothesis testing as well as instantaneous trend analysis for March
4. glucose_2021_analysis.ipynb: Notebook that contains trend exploration from January to April 2021 for anomaly detection

---

Screenshots

Refer to graphs folder for all images

1. Summary Data Analysis

---

2. 2021 Glucose Level Outlier Detection

• 2021 Glucose Trends

• 2021 Outlier Detection Using Isolation Forest

---

3. March Blood Sugar Analysis

• Examining Exercise vs Low Glucose Levels at Night

• Amount of Exercise Violin Plots

• Instantaneous Blood Sugar Trend Analysis

Coming Soon