Explainable Machine Learning for Nomophobia Prediction: Insights through SHAP Analysis

The fear or anxiety of not having access to a mobile phone or its services, known as nomophobia, is becoming more and more common and has serious social and psychological repercussions. This study uses Explainable Artificial Intelligence (XAI) method in conjunction with machine learning (ML) classifiers such as Naive Bayes (NB), K-Nearest Neighbour (KNN), and Support Vector Machine (SVM) to categorise nomophobia severity based on data from medical surveys. ML models were trained on dataset collected from various populations. $k$-Fold cross-validation was used to assess the model's performance with evaluation metrics such as accuracy, recall, precision, $F_{1}$, and Geometric-Mean (GM). Multi-Criteria Decision Analysis (MCDA) was conducted using well-known Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) method, and SVM stands out to be best among three models. Shapely Addictive Explanations (SHAP), an XAI technique, were used to interpret the results, offering information on the significance of each feature and the reasons behind each prediction. SHAP showed excellent predictive performance with perfect accuracy metrics. These results support better clinical decision-making by offering a transparent and efficient method for assessing nomophobia risk with trust.