Abstract:
"This research paper presents an activity recommendation system for mental stress using machine
learning techniques. Mental stress is a significant public health concern that can lead to various
mental and physical health problems. To alleviate mental stress, regular physical activity is
recommended. However, people often struggle to find suitable activities that can help them
manage their stress levels.
To solve this problem, an activity recommendation system is proposed that uses machine learning
models to predict users stress and provide personalized recommendations. Two machine learning
models were used in the research: Naive Bayes Multinomial for activity prediction and contentbased filtering using cosine similarity for recommendation. The Naive Bayes Multinomial model
was used to predict the probability of the user’s stress. The content-based filtering model used
Tfidf vectorized cosine similarity to recommend activities that can help manage mental stress
levels.
The proposed system was evaluated, and the results showed that the system can effectively predict
users' stress and provide relevant recommendations for managing mental stress. The system's
performance was evaluated using various metrics such as precision, recall, and F1-score, and the
results showed that the prediction system outperformed the baseline models. The recommendation
was evaluated using the MAP for predictions as the output of the recommendation system is a
range and it depends on different features.
Overall, the proposed activity recommendation system for mental stress has the potential to
improve users' mental well-being by providing personalized and relevant activity
recommendations. The system can be used by mental health professionals to prescribe suitable
activities to their patients or by individuals who want to manage their stress levels. The research
demonstrates the effectiveness of machine learning techniques in developing personalized activity
recommendation systems for mental health."
