HairSentinel: a time-aware anomaly detection framework for forecasting hairfall trends using temporal fusion transformers
Year: 2025
Type: article
Abstract: Hairfall is a primary concern for many individuals worldwide today. Hair strands may fall due to various conditions such as hereditary factors, scalp health issues, nutritional deficiencies, hormonal fluctuations, or irregular sleep cycles. Our study presents a novel approach to detecting hairfall trends over time. While traditional methods infer hairfall rates using CNN and SVM models-classifying types of hairfall based on high-resolution images and complex techniques-this study addresses the issue by analyzing user-provided data through simple, straightforward questions, maintaining ease of use. Each attribute is collected using a time-centric approach on a daily or weekly basis. For time series anomaly detection, we utilize LSTM, Random Forest, and the Temporal Fusion Transformer (TFT) to model hairfall fluctuations and compare them with the ARIMAX model across various metrics to identify the most suitable one. The TFT model is selected as the most suitable, with 97.5% accuracy and 97.4% precision over other models supporting anomaly detection. This allows us to establish normal margins of deviation from typical hair shedding cycles. This study enables the proactive detection of anomalies, indicating sudden increases or decreases in hairfall due to hormonal fluctuations. The results support the early identification of potential health risks before they become intensified and help suggest appropriate dietary plans.
Institution: Vellore Institute of Technology University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,353
Funder: VIT University
OpenAlex ID: W4415170595