Open AccessRecognition of Forest Fire Spruce Type Tagging using Machine Learning Classification
Author(s) -
M. Shyamala Devi,
Shefali Dewangan,
Satwat Kumar Ambastha,
Anjali Jaiswal,
Sairam Kondapalli
Publication year - 2019
Publication title -
international journal of recent technology and engineering
Language(s) - English
Resource type - Journals
ISSN - 2277-3878
DOI - 10.35940/ijrte.c5176.098319
Subject(s) - random forest , naive bayes classifier , support vector machine , computer science , decision tree , damages , machine learning , feature (linguistics) , artificial intelligence , environmental science , linguistics , philosophy , political science , law
In recent times, the natural resources are demolished due to the technological growth. The agricultural and the forest area are transformed to industries, Storage Warehouse and Container logistics companies to facilitate the living standards. This leads to scarcity of natural resources for the people to live a comfortable life. Due to the change of natural environment and fluctuations in the climate conditions, the forest has the chance of occurrence of fire. The forest fire is the resultant of high temperature, land mine, flight crashes and satellite damages from the environment. The precaution must be taken in advance to protect the coverage of fire. The less attention to fire control may lead to entire damage of the forest and the spreading of fire occurs due to the high wind blow. This makes researchers to focus on helping the forest area to overcome from the fire attack. The detection of fire type is a challenging task after the occurrence of the damage. With this view, we address the prediction of fire type classification using machine learning classification algorithms. The Forest Cover Type dataset is downloaded from UCI Data warehouse repository and done with classification analysis. The prediction of absent hours is achieved in the methodology of four steps. At first, the important feature attributes are found and depicted as a chart. Secondly, the raw dataset is applied to all the classification models like Logistic regression, Kernel SVM, KNN, Decision Tree, Naïve Bayes and Random Forest. Thirdly, the dataset is reduced with PCA and then the reduced dataset is fitted to all the classifiers. Fourth, Performance analysis is done by analyzing the performance metrics like Accuracy, FScore, Recall and Precision. The real time execution is performed by python in Anaconda Spyder Navigator Integrated Development Environment. Experimental Result shows that the Random Forest classifier is obtained with the accuracy of 92% before applying PCA. After applying PCA, the classifier namely Random forest is analyzed to be having the accuracy of 78% for 15 components, 83% for 20 components and 89% for 25 components.