Mean Squared Error And Regression Analysis

893 Words 4 Pages
Performance Evaluation
In order to assess the model accuracy, it is necessary to use some quantitative measures of learning. In this study the Mean Squared Error (MSE) and regression analysis were used to evaluate the model performance. MSE is a useful measure of success for numeric prediction and is calculated using Eq. (1). It is worth mentioning that small values of MSE indicate better performance of the ANN model. It was found that the optimum performance of the model is at 25 neurons with MSE 0.1481. The accuracy of the model with different neurons is shown in Table 5.

MSE=(∑_(j=0)^P▒∑_(i=0)^N▒〖(t_ij-y_ij)〗^2 )/NP (5)

Where P is the number of output possessing elements, N is the number of observations, tij are the target outputs and yij are the actu9al outputs.
TABLE 6 accuracy for the network
Neurons in the Hidden Layers MSE Values
12 0.3263
20 0.1652
25 0.2672
28 0.2870
32 0.6301

Fig. 11 shows the regression analysis of targets and outputs for Levenberg-Marquardt algorithm during training and testing process. The best fit lines in Fig. 10 demonstrate the relationship between the desired value and actual value. As shown in Fig. 10 the R value of ANN model is 0.8134 which indicates that LM algorithm has
…show more content…
One project was selected for the case study. One of the problems of the selected project is that the techniques selection procedures are based on elicitor experience. The aim of this case study is to check whether the proposed ANN model predicts more effective techniques than the current requirements engineering methods. The contextual situation of the selected project, that is, the particular values for the influential attributes is shown in Table 7. Table 8 shows the normalized values calculated using Eq. (4) for each attribute. These normalized values are considered the input data for the ANN

Related Documents