| Autor: |
El Houcine Sebbar, Mohamed Choukri, Souad Bechrouri, Abdelilah Monir, Hamid Mraoui, Ennouamane Saalaoui |
| Rok vydání: |
2019 |
| Předmět: |
|
| Zdroj: |
Proceedings of the New Challenges in Data Sciences: Acts of the Second Conference of the Moroccan Classification Society. |
| DOI: |
10.1145/3314074.3314076 |
| Popis: |
Strong demand for vitamin D diagnosis was observed in recent years. This trend has led to an increase in public health expenditures as well as for the patient. The vitamin D prediction was an alternative for saving patients the extra charge of blood tests. We compared some statistical methods in order to predict the vitamin D levels based solely on biochemical parameters, age, and sex. A set of hospitalized patients from different departments of the University Hospital Centre of Oujda and having a valid value for vitamin D and various biochemical parameters were included. There were 124 patients aged between 9 to 87 years old (mean = 45.19, median = 49) and 17 variables. Vitamin D was predicted using linear regression, Multivariable Adaptive Regression Spline (MARS), Random Forest (RF) and Support Vector Regression (SVR). Two statistics were used to compare the different models: Root Mean Squared Error (RMSE) and the Mean Absolute Error (MAE). The test correlation was showed weak correlations between vitamin D and some biochemical parameters such as calcium and glucose. These comparisons were demonstrated the SVR model performed better than random forests and MARS in the case of a small size database. This prediction may help to identify patients with a real risk of vitamin D deficiency. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
