Machine learning based model for detecting depression during Covid-19 crisis.

Area of Science:

• Mental Health Research
• Computational Psychiatry
• Public Health

Background:

• The COVID-19 pandemic has exacerbated mental health issues globally, with depression identified as a significant public health concern.
• Depression is linked to increased risks of other health conditions, suicide, and premature death, highlighting the need for early detection.
• Conventional methods for assessing depression can be time-consuming and subjective.

Purpose of the Study:

• To investigate the efficacy of machine learning models for early depression detection.
• To compare the performance of Decision Tree, KNN, and Naive Bayes algorithms in analyzing depression survey data.
• To propose a machine learning-based approach for more efficient depression assessment.

Main Methods:

• A survey comprising 21 questions, based on the Hamilton Depression Rating Scale and psychiatric consultation, was administered.
• Python's scientific programming and machine learning techniques (Decision Tree, KNN, Naive Bayes) were employed for data analysis.
• Performance metrics including accuracy and latency were used to compare the selected machine learning models.

Main Results:

• K-Nearest Neighbors (KNN) demonstrated superior accuracy in depression detection compared to other evaluated methods.
• Decision Tree exhibited better performance in terms of detection latency, offering faster identification of depression.
• The study provides a quantitative comparison of machine learning algorithms for depression screening.

Conclusions:

• Machine learning models offer a viable and potentially more efficient alternative to traditional depression assessment methods.
• KNN is recommended for its high accuracy in identifying depression, while Decision Tree is suitable for rapid screening.
• The development of a machine learning-based model can significantly improve the early detection and intervention of depression.