An Explainable Neuro-Fuzzy Model for Hypertension Prediction

Description

This dataset consists of 1249 clinical records created for research on explainable machine learning and neuro-fuzzy modelling for hypertension prediction. The data set combines demographic, anthropometric, clinical, lifestyle and family history data that are relevant to hypertension risk classification. The data set includes 14 variables: Age, Gender, Height (cm), Weight (kg), Body Mass Index (BMI), Systolic Blood Pressure, Diastolic Blood Pressure, Cholesterol (mg/dL), Blood Glucose (mg/dL), Smoking, Alcohol Intake, Physical Activity, Family History, and Hypertension Status. The Hypertension_Status variable is the target variable for hypertension classification. The data set was prepared for the study “An Explainable Neuro-Fuzzy Model for Hypertension Prediction” at Osun State University, Osogbo, Nigeria. The accompanying research methodology is based on the selection of clinical variables for model development, which are Age, BMI, Systolic Blood Pressure, Diastolic Blood Pressure, Cholesterol Level, and Blood Glucose Level. The proposed modelling approach is a combination of fuzzy reasoning and machine learning to deal with the nonlinear relationships and uncertainty in clinical prediction. The research methodology involves data preprocessing, categorical encoding, feature selection, feature scaling, partitioning of the dataset, development of a neuro-fuzzy model, baseline machine learning modelling, performance evaluation, and explainability analysis using SHAP. The proposed neuro-fuzzy system is based on Sugeno-type fuzzy inference system with linguistic variables, membership functions and expert defined IF–THEN rules. The data set is designed to facilitate academic research, experimentation, benchmarking, machine learning development, explainable artificial intelligence research, and investigation of intelligent approaches for hypertension risk prediction. It is not intended to be used as a clinically validated diagnostic tool and predictions made by models trained on this data should not be used as a substitute for clinical assessment or clinical decision making.

Steps to reproduce

(1) Download Hypertension Dataset.xlsx from the Mendeley Data repository. (2) Open the UTH DATASET worksheet using Microsoft Excel, Python/Pandas, R, MATLAB, or another software capable of processing Excel files. (3) Inspect the dataset and identify the 14 variables: Age, Gender, Height_cm, Weight_kg, BMI, Systolic_BP, Diastolic_BP, Cholesterol_mg_dL, Blood_Glucose_mg_dL, Smoking, Alcohol_Intake, Physical_Activity, Family_History, and Hypertension_Status. (4) Use Hypertension_Status as the target variable for hypertension classification. (5) Perform data-quality checks for missing values, duplicate records, inconsistent entries, and inappropriate values. (6) Encode categorical variables such as Gender, Smoking, Alcohol Intake, Physical Activity, Family History, and Hypertension Status into numerical representations where required. The project methodology specifies the use of Label Encoder for categorical variables. (7) Select the clinical predictors required for the modelling experiment. The project identifies Age, BMI, Systolic Blood Pressure, Diastolic Blood Pressure, Cholesterol Level, and Blood Glucose Level as the selected input variables. (8) Apply feature scaling/standardization to the selected numerical variables. The project specifies StandardScaler for transforming the features to zero mean and unit variance. (9) Partition the processed dataset into training and testing subsets using an 80:20 ratio. A random state of 42 is specified to support reproducibility. (10) Develop the neuro-fuzzy model using the Simpful fuzzy inference library in Python. The proposed model uses a Sugeno-type fuzzy inference system. (11) Define triangular membership functions for the input variables using linguistic categories such as Low, Medium, and High. (12) Construct the fuzzy rule base using IF–THEN rules representing relationships among hypertension risk factors. (13) Generate the neuro-fuzzy prediction using Sugeno inference and convert the resulting output into the required hypertension classification using the specified threshold of 0.5. (14) Develop baseline machine learning models, including Logistic Regression, Support Vector Machine (SVM), and Random Forest Classifier, using the same training and testing datasets for comparison. (15) Evaluate the models using appropriate classification metrics, including Accuracy, Precision, Recall/Sensitivity, F1-score, and classification reports. (16) Apply SHAP-based explainability analysis to investigate the contribution of clinical features to model predictions. The project specifies SHAP summary and individual prediction explanations. (17) Compare the performance of the neuro-fuzzy model with the baseline machine learning models and report the resulting predictive and explainability findings

Institutions

• Osun State University

  Osun State, Osogbo

Categories

Licence