The present health scenario indicates that thyroid diseases are a common
challenge experienced by most individuals. According to the statistics in India,
one out of eight women suffer from thyroid-related conditions. Hyperthyroid,
hypothyroid, or thyroid cancer are categories of thyroid disorder. It is imperative
to maintain optimum levels of secretion of the thyroid hormones as the imbalance
could lead to thyroid diseases. Therefore, thyroid patients must be vigilant
regarding their iodine intake and follow a customized daily diet and exercise plan.
The diet plan, along with balanced iodine levels, must also be able to meet the
patient's nutritional needs. A personalized diet plan could help thyroid patients to
be more aware and focused on their body metabolism. Existing recommender
systems usually provide generic diet recommendations, and unfortunately, it may
not be beneficial to patients suffering from a specific disease.
Content-based Neighborhood-Conditional RBM (CB-NCRBM) model has
posited to recommend Top-3 diet and exercise plans for thyroid patients. The
proposed model considers the joint probability distribution of different scores
using the user profile. Similarly, preference and health scores are estimated based
on content features. The model feeds these scores as visible units to conditional
RBM. The proposed model also integrates several content-based features such as
users' physiological profiles, thyroid disease information, food, and exercise
preferences. The proposed recommender model validates the experimental results
using recommendation error and classification accuracy metrics. The proposed
hybrid model outperforms several popularly used recommendation models, such
as collaborative filtering, content-based, and pure RBM models. The system also
provides a feedback loop to enhance the quality of the recommended diet and
exercise plans based on user experience.
