Enhancing the Accuracy in Predicting Snow Avalanche with Support Vector Machine Algorithm Compared with K-Nearest Neighbour

To quantify the Predicting Snow Avalanche using Support Vector's machine learning algorithm and K-nearest neighbor classifiers. The Snow Avalanche datasets comprises 2400 analyses with 215 rows and 15 columns for the characteristics of Snow Avalanche as displayed in Table 3 above The forecast on the occurrence of Strokes is made through Support vector machine (N=10) and K-Nearest Neighbor (N=10) as a way of enhancing the accuracy of the forecast. Sample size: If groups for this study are considered ClinCalc calculates a sample size of 10 for each group with G Power 0. 8 and an alpha of α = 0. 05. Precision to predict Stroke is measured for the assessment of this disease, as it is stated in the equation above. In comparing the accuracy of the models that have been implemented for the Snow avalanche prediction, Support Vector Machine performs better with 97. 89% compared to K-Nearest Neighbor model with 77. 89%. This implies statistical significance. When comparing the average accuracy of mean, it can be seen that SVM has achieved better results compared to K-NN algorithm. While the Support Vector Machine has a better average accuracy of 90. 95% compared to the 84. 43% from the Random Forest Algorithm, analyzing the dispersion density shows that the Random Forest Algorithm has a higher standard deviation of 7. 86, and standard mean error of 7. 57 compared to the Support Vector machine's 3. 02, and 2. 94 respectively.

I. Introduction

One of the most dangerous natural phenomena observed worldwide is a snow avalanche; mountainous territories being most endangered. Snow avalanches, as any other kind of strokes, lead to deaths and disabilities and pose a considerable threat to inhabitants. Therefore, predication of snow avalanches is a key factor in the reduction of these risks, particularly in areas that have high occurrence [2] Globally, in 2019 snow avalanche incidents was estimated to be in the neighborhood of resulting to numerous fatalities and property damage. Present methods of managing the risk of an avalanche pay much attention to the ability to predict an avalanche, as it is comparable with managing the risk of cardiovascular diseases and strokes [2]. Trauma can be averted by early management and control of risk factors that can be eliminated resulting in a decrease in the number of avalanche occurrences. The possibility of scaling up avalanche prediction algorithms has the opportunity of positively impacting a significant number of public safety programs [3]. Emergency response and government organizations may also use this information for establishing specific awareness creation, educational and community mobilization efforts toward the prevention of snow avalanche incidence. AI and ML technologies have enhanced the prediction of avalanche hazards as it has advanced the speed and acuarcy of models. [4] ML techniques can process large sets of information about the terrain, weather, and prior accidents, if any.