機器學習於臺灣數位近用與網路使用行為之研究

A Study on Machine Learning Applications in Digital Access Survey and Network Usage Behavior in Taiwan

本研究以調查研究資料庫（Survey Research Data Archive，SRDA）上「113年數位近用調查」為資料來源，結合「數位科技近用模式」為主軸，並運用機器學習技術，以集群分析、羅吉斯迴歸、決策樹、隨機森林、類神經網路及簡易貝氏分類等演算法，探討「數位近用條件與網路使用行為的關聯」及「數位近用與使用行為如何與主觀幸福感（生活滿意度、科技適應度）互相連動」等議題，透過機器學習方法將「數位參與品質」視為社會永續的重要面向，補足過往僅以基礎網路環境建置與否作為指標的框架，並以「動機、物質、技能、使用」四層環節為基礎提出具體實務建議以協助發掘優化數位能力與服務設計方向，縮減不同群體間的數位落差，並促進全民普遍且具意義的數位參與，作為邁向包容性、永續性數位社會的依據。

This study uses the“2024 Digital Access Survey”(Survey Research Data Archive, SRDA) as its data source, and is grounded in the“Digital Access Model.”By applying machine learning techniques - including Cluster Analysis, Logistic Regression, Decision Trees, Random Forests, Neural Networks and Naïve Bayes Classifier - this study examines issues such as '' the relationship between conditions of digital access and patterns of internet use '', and '' how digital access and usage behaviors interact with subjective well-being (life satisfaction and technological adaptability) '' .
By treating“the quality of digital participation”as a key dimension of social sustainability, using machine learning methods, which supplements the previous framework that only used the establishment of basic network environment as an indicator, this study aims to provide concrete recommendations based on four layers, motivation, physical access to devices, digital skills, and actual usage. The goal is to help narrow digital gaps between different social groups, promote universal and meaningful digital participation, and offer an evidence base for building an inclusive and sustainable digital society.