Implementasi Teknik Data Mining untuk mendeteksi Gangguan Psikologis Pasca Melahirkan

Muhammad Zulfadhilah
Putri Yuliantie

Tematik : Jurnal Teknologi Informasi Komunikasi
Vol 10 No 2 (2023): Tematik : Jurnal Teknologi Informasi Komunikasi (e-Journal) - Desember 2023

Currently, in the health sector, various studies are conducted, one of which is in terms of psychological health. One of the psychological disorders occurs in postpartum mothers, around 10-15% of postpartum mothers experience psychological disorders such as anxiety and depression, the high number is a concern in society, especially families. This research was conducted to assist the government in supporting research priorities on health independence using current technology. This is also a research urgency, namely to minimize psychological disorders that occur in postpartum mothers. One of the problem-solving approaches proposed in this research is to use technology with the implementation of Data Mining which has the advantage of predicting the likelihood of a person having certain diseases or health disorders. Data Mining is an approach that is often used and has been used as a reference in health nursing by using the results in two branches, namely for decision support and policy making. Implementation of the Data Mining algorithm provides exposure to analyze, detect, and predict the presence of disease and assist doctors in making decisions with early detection and appropriate management. One type of data mining is using Naive Bayes. From the results of the model evaluation, it can be concluded that the Naive Bayes model shows good performance in detecting postpartum psychological disorders. Evaluation values describe the model's ability to distinguish between positive and negative classes, with an AUC value of around 0.878. The model's accuracy of about 0.819 indicates its ability to correctly predict about 81.9% of the cases tested. F1-Score and Recall around 0.817 and 0.819 respectively indicate a balance between positive prediction and positive instance-finding ability. The Confusion Matrix also describes the performance of the model. Despite having a significant number of True Positives (TP) (870), the number of False Positives (FP) is noteworthy (162), indicating some false positive predictions.