Recipe Recommendation for the Elderly Based on Collaborative Filtering Algorithm and Neural Network
Because there is no system developed for the characteristics of the elderly in the application of food recommendation at this stage, many nursing homes can only cook for the elderly according to the experience of dietitians, lacking diversity. This project obtains some data from the background of th...
Uloženo v:
| Vydáno v: | 2024 IEEE 2nd International Conference on Sensors, Electronics and Computer Engineering (ICSECE) s. 660 - 663 |
|---|---|
| Hlavní autor: | |
| Médium: | Konferenční příspěvek |
| Jazyk: | angličtina |
| Vydáno: |
IEEE
29.08.2024
|
| Témata: | |
| On-line přístup: | Získat plný text |
| Tagy: |
Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
|
| Shrnutí: | Because there is no system developed for the characteristics of the elderly in the application of food recommendation at this stage, many nursing homes can only cook for the elderly according to the experience of dietitians, lacking diversity. This project obtains some data from the background of the Chinese small program "Yikang", which is used by elderly people and nutritionists in nursing homes to score dishes, and develops a collaborative filtering neural network. In order to further improve the cold start of common collaborative filtering algorithm and the insufficient ability to process sparse matrix, the algorithm part uses the pre-trained Neural Collaborative Filtering (NCF) network. First, the matrix decomposition part of the common collaborative filtering algorithm is replaced by a deep neural network. The user and dish features are expressed as hidden vectors, and then the interaction function is learned from the data. Then the gradient descent method is used to solve the problem of sparse matrix processing. The model output is normalized to make the output probability within the range of [0], [1]. This paper adds multiple hidden layers between the input layer and the output layer of the neural network and introduce the activation function to transform the hidden variables. In order to further fit the optimization target, the forward propagation and backward propagation are used to compare the difference between the predicted value and the actual value and adjust the parameters. After adjusting the parameters of the improved algorithm, the accuracy of the same dataset is improved from 70% to 84% compared with the original collaborative filtering algorithm. The accuracy of prediction results has been greatly improved. Therefore, it can be proved that the NCF network can effectively combine the physical condition and taste preferences of the elderly and recommend dishes with greater accuracy than the general recommendation procedure. |
|---|---|
| DOI: | 10.1109/ICSECE61636.2024.10729407 |