Prediction of Bone Loss Rate Based on Oesteoporosis Risk Features

Osteoporosis is found from the set of electronic health care data and the bone loss rate is calculated from the available osteoporosis risk feature sets. The main disadvantage arise in existing work is that, bone loss rate cannot be predicted accurately due to limited number of features. This is resolved in our proposed work by introducing the relevance based feature selection which helps to predict the related features with less availability of feature. Osteoporosis measured by bone mineral density, bone fracture risk is determined by the bone loss rate and various factors such as family history and life style. . The Deep belief network is used for fine tuning of risk factors. In this learning process, two stages of process are carried out. They include pre-training and fine tuning. In the pre-training phase, most important risk factors with model parameters are used to calculate contrastive divergence and it minimizes the record size. In the fine tuning phase comparison is made with the results achieved in the previous phase with the ground truth value g1 and again the same comparison done with ground truth value g2, were g1 is refer to as osteoporosis and g2 is refer to as a bone loss rate. The final results are applied to confusion matrix to describe the performance of classification model based on the comparison results, the following are calculated Accuracy, Precisions, Recall, and F-Measure.