Predicting Noise-Induced Hearing Loss among Elderly Residents Near Powerloom Industries Using Machine Learning

Background: Environmental noise from small-scale industries, particularly powerloom clusters, is an underrecognized public health concern in India. Older adults in these settings are especially vulnerable due to age-related auditory decline compounded by chronic noise exposure. With expanding semi-urban industrialization and a growing elderly population, noise-induced hearing loss (NIHL) is emerging as a significant yet overlooked health burden. This study estimated the prevalence of NIHL among elderly residents near powerloom industries and evaluated key predictors and machine learning models for community-level screening. Methodology: A community-based cross-sectional study was conducted in Kumarapalayam, Tamil Nadu, among 436 adults aged ≥60 years. Participants were categorized into an exposed group (n = 218; residing <500 m from powerloom units) and a control group (n = 218; residing >2 km away). Environmental noise levels were recorded using standardized sound level meter, showing substantially higher mean daytime noise exposure among the exposed group (77.6 ± 5.67 dB) compared to the control group (52.35 ± 3.95 dB). Hearing thresholds were assessed using validated mobile audiometry. Four ML classification models Random Forest, Support Vector Machine (SVM), k-Nearest Neighbor (KNN), and Logistic Regression were trained and evaluated to predict NIHL from demographic and exposure-related variables. Results: Bilateral hearing loss was markedly higher in the exposed group (65.14%) than in the control group (35.18%). Random Forest demonstrated the strongest performance, achieving an accuracy of 93.4%, a precision of 93.0%, and a recall of 93.2%, outperforming the other models. Predictive variables such as age, proximity to powerloom units, duration of residence, and measured environmental noise levels played significant roles in model performance. Conclusions: Elderly individuals residing near powerloom industries experience significantly greater noise exposure and a correspondingly higher prevalence of NIHL. Machine learning demonstrates strong potential as a practical, field-friendly tool for early identification of at-risk individuals in resource-limited settings.