Prediction of smartphone overdependence and analysis of its influencing factors among older adults based on machine learning.

Abstract

BACKGROUND: With the widespread use of smartphones among middle-aged and older adults, the risks associated with excessive use and dependence on smartphones have become increasingly apparent. This study aims to identify and predict the risk factors for smartphone overdependence among older adults in South Korea, utilizing machine learning methods to construct predictive models. METHODS: We utilized panel data from the "2023 Smartphone Overdependence Survey" provided by the National Information Society Agency (NIA) of South Korea. This study specifically focuses on the older adult population aged 60 and above, identifying key factors influencing their smartphone overdependence. A variety of machine learning-based binary classifiers were evaluated, including XGBoost, SVM, LR, KNN, DT, and NB. Their predictive accuracy and performance were compared comprehensively. Model performance was assessed using multiple metrics, including confusion matrix, accuracy, precision, recall, F1 score, and AUC. RESULTS: The XGBoost classifier performed the best in predicting smartphone overdependence among older adults, with an accuracy of 0.925. Through feature importance analysis, we found that demographic characteristics, time composition of smartphone use, awareness of smartphone overdependence problem, and content of smartphone use were the main influencing factors in predicting smartphone overdependence among older adults. CONCLUSION: Artificial intelligence algorithms have the potential for predictive and explanatory capabilities, identifying the risk of smartphone overdependence among older adults and the associated risk factors. This has significant theoretical and practical implications for understanding and addressing this issue.

AI evidence extraction

Main findings

Using 2023 Smartphone Overdependence Survey panel data from South Korea, several machine-learning binary classifiers were compared to predict smartphone overdependence in adults aged 60+. XGBoost performed best (accuracy 0.925), and feature-importance analysis indicated demographic characteristics, time composition of smartphone use, awareness of the overdependence problem, and content of smartphone use were key influencing factors.

Outcomes measured

• smartphone overdependence
• predictive model performance (accuracy, precision, recall, F1 score, AUC)

{
    "study_type": "other",
    "exposure": {
        "band": null,
        "source": "smartphone",
        "frequency_mhz": null,
        "sar_wkg": null,
        "duration": null
    },
    "population": "Older adults aged 60 and above in South Korea",
    "sample_size": null,
    "outcomes": [
        "smartphone overdependence",
        "predictive model performance (accuracy, precision, recall, F1 score, AUC)"
    ],
    "main_findings": "Using 2023 Smartphone Overdependence Survey panel data from South Korea, several machine-learning binary classifiers were compared to predict smartphone overdependence in adults aged 60+. XGBoost performed best (accuracy 0.925), and feature-importance analysis indicated demographic characteristics, time composition of smartphone use, awareness of the overdependence problem, and content of smartphone use were key influencing factors.",
    "effect_direction": "unclear",
    "limitations": [],
    "evidence_strength": "insufficient",
    "confidence": 0.7399999999999999911182158029987476766109466552734375,
    "peer_reviewed_likely": "yes",
    "keywords": [
        "smartphone overdependence",
        "older adults",
        "South Korea",
        "machine learning",
        "XGBoost",
        "survey",
        "risk factors",
        "prediction"
    ],
    "suggested_hubs": []
}

Comments

Log in to comment.

No comments yet.