Monitoring data quality using ANN-based Hybrid Exponentially Weighted Moving Average control charts

Abstract

Background: Quality is essential to determining public health and is required for the healthcare industry to sustain itself. An important component of public health is air quality. This research endeavour aims to monitor and enhance air quality by creating control charts with Artificial Neural Networks (ANN), a machine-learning technology based on the human brain. Objective: The objective of this paper is to evaluate the effectiveness of the Shewhart individual X bar chart, the ANN-based Hybrid Exponentially Weighted moving average (HEWMA), and the ANN-based Exponentially Weighted Moving Average (EWMA) for Delhi's air quality index data from January to December 2023. Methods: According to the most current studies, only very few studies used the hybrid EWMA charts with ANNs. One of the most effective ways to activate a new hybrid EWMA (HEWMA) control chart and monitor the process mean is to combine two EWMA control charts. ANN-based HEWMA system was one approach that demonstrated potential for air quality index variable detection. After that ANN-based EWMA control charts and Shewhart X bar control charts are developed. The Average Run Length (ARL) and number of outliers were determined for all three charts. Findings: When comparing ARL values, the number of outliers, and false alarms the results suggested that ANN-based hybrid exponentially weighted moving average provides better outcomes than alternative charts.ANN-based HEWMA control has a very good ability to detect out-of-control zones. This study discovered that the suggested HEWMA control chart is more effectively successful in detecting process mean points than the EWMA control chart. Aspects of ANN-based HEWMA can provide faster identification in comparison to other charts. Novelty: This work combines EWMA with ANN in an innovative method to develop the ANN-based HEWMA control chart. Applying this hybrid chart to standard charts improves the detection of process anomalies, decreases outliers and false alarms, and increases sensitivity and accuracy. It significantly increases the use of SPC charts because to its faster and more reliable monitoring capabilities, particularly in the healthcare sector.