E-Prediction of Water pH Meter Calibration Equipment Feasibility at Public Health Laboratory Using the Decision Tree Method

Abstract

Abstract: Digital transformation in the public health laboratory sector requires the optimal utilization of calibration data to support more effective and objective decision-making processes. The Public Health Laboratory (LABKESMAS) still experiences limitations in managing water pH meter calibration data, where the feasibility assessment of laboratory equipment is generally conducted manually using Microsoft Excel calculations. This process is prone to human error, data inconsistency, and inefficient evaluation procedures. Therefore, this study aims to analyze and implement the Decision Tree algorithm in a web-based E-Prediction system to determine the feasibility of water pH meter calibration equipment based on historical and real-time calibration data. The research methodology applied in this study uses a quantitative approach with data mining techniques through the Knowledge Discovery in Database (KDD) stages, including data collection, data preprocessing, Decision Tree implementation, and interpretation of classification results. The dataset used consists of historical calibration records and newly input calibration data of water pH meters measured using standard buffer parameters pH 4, pH 7, and pH 10. The preprocessing stage includes data cleaning, validation, and deviation calculation between measured values and standard values. The classification process utilizes predefined deviation threshold rules embedded within the Decision Tree structure.

The findings indicate that the Decision Tree algorithm effectively classifies equipment feasibility into three categories: feasible, requiring recalibration, and unfeasible. Equipment with deviation values less than or equal to 0.02 pH is categorized as feasible, deviations between 0.02 pH and 0.05 pH are classified as requiring recalibration, and deviations greater than or equal to 0.05 pH are categorized as unfeasible. The developed web-based E-Prediction system achieved 100% classification accuracy according to the implemented logical rules. In addition, the system improves operational efficiency, minimizes manual calculation errors, accelerates the decision-making process, and provides automatic PDF-based report generation. Consequently, the implementation of the Decision Tree algorithm has proven effective in supporting transparent, consistent, and data-driven laboratory equipment feasibility evaluation processes at the Public Health Laboratory.