Please check the GitHub link for full reproducible code and dataset: https://github.com/nipunkalraa/Predicting-Food-Consumption-using-Machine-Learning_UK/tree/main

This project uses machine learning techniques to predict whether individuals are likely to consume above-average amounts of red meat based on socio-demographic characteristics. The analysis helps identify key demographic factors that can inform targeted marketing strategies.

📊 Dataset

• Individual-level survey data containing:

• Demographic variables (Age, Sex, Ethnicity, Health status, Employment)

• Food consumption patterns (including red meat, alcohol, food and vegetables, fish consumption)

📦 Packages Used

• tidyverse: Data transformation and visualisation

• caret: Machine learning algorithms and model evaluation

• colorspace: Colour palette manipulation for visualisations

• ggplot2: data visualisation

🔧 Methods

Data Preprocessing:

• Removed extreme age groups (0-15 and 75+)

• Encoded categorical variables (One-Hot for Sex/Work, Ordinal for Age/Health)

• Created binary outcome variable for red meat consumption (above/below average)

Machine Learning Models:

• Logistic Regression: Baseline binary classification

• Random Forest: A more complex ensemble learning approach

• Used 80/20 train/test split with 10-fold cross-validation

Model Evaluation:

• Accuracy, Kappa, Sensitivity, Specificity

• Precision, Recall, F1 Score

• Variable importance assessment

🔍 Key Findings

The most important predictors for red meat consumption: Sex (Male)

• Ethnicity (White)

• Employment status (employed)

• Age groups (45-64, 65-74)

The Random Forest model outperformed Logistic Regression across performance metrics