Simulations from Support vector mac

Simulations from Support vector machine (SVM) and K-nearest neighbour method is compared with respect to observed data by the help of box plots and cumulative distribution functions for both historical period and future projections at all five locations. Box plots for historical period reveals that K-nn method gives results close to the observed one while SVM sometimes underestimates and overestimates the results and also unable to predict peak values. Figure 3 shows Box-plot and Figure 4 shows CDF at one of the five locations. KOLMOGOROV- SMIRNOV TEST, a non parametric test, which is used to compare two sample data sets (weather it belongs to the same distribution or not) is also conducted. The K-S test is conducted first on observed data with simulation from SVM model and second on observed data with simulations from KNN model. Both reject the null hypothesis which signifies that it belongs to the same family of distribution.