Evaluation of parents' digital parenting self-efficacy through data mining

In this study, it was aimed to evaluate the digital parenting self-efficacy of secondary school students' parents with data mining in terms of certain independent variables. First, parents were profiled according to their digital parenting self-efficacy status. Then, the patterns among the parents' answers were discussed, and the predictive performance of the dependent variable (digital parenting self-efficacy) on the independent variables was presented. Suggestions were presented for parents to raise their children while being aware of digital risks and opportunities. The participants of the study consisted of 597 parents of students in secondary schools in a city district in eastern Turkey. However, due to the missing data in the data set, a total of 370 parents' data used for analysis purposes. "Personal Information Form" and "Digital Parenting Self-Efficacy Scale" were used to collect data. Descriptive statistics, independent sample t-test, and one-way analysis of variance for independent groups were used to examine the significant differences of parents' digital parenting self-efficacy according to different variables. In addition, the association rule of data mining, prediction/classification, clustering analysis, and feature selection method were used in the analysis. According to the results, the profiles of the parents were divided into three clusters through the clustering method. It was observed that although the total digital parenting self-efficacy score of the parents was low, they seemed to be sensitive in warning their children about in-game purchases and internet shopping, and parents with high digital parenting self-efficacy scores were more sensitive than other parents in configuring privacy/security settings. According to the association rule analysis results, it was concluded that parents used social media actively, but they were unable to install security software to protect themselves from social media risks. In addition, it was found that while parents who used the internet for research and who always used their smart phones were sensitive to warn and inform their children about e-commerce, and parents who never used a smart watch/wristband and computer were inadequate in terms of installing security software. As a result of the estimation analysis, it was determined that the estimation performance of digital parenting self-efficacy scores was 76.8% using the Random Forest algorithm, using all the independent variables in the study. It was concluded that the independent variables determined by the feature selection method were 72% estimation success with the Random Forest algorithm. By applying the feature selection method on the independent variables in the study, important variables on digital parenting self-efficacy were determined. These variables were father's occupation, mother's education level, number of children, communicating with the child about the time spent on the Internet, daily Internet use of the child, frequency of desktop computer use, and frequency of use of smart watches/wristbands. According to the results of the one-way analysis of variance, there was a significant difference between digital parenting self-efficacy scores and these variables.