SVM integrated case based restarting GA for further improving solar flare prediction

• Published in: 2016 IEEE International Conference on Systems, Man, and Cybernetics (SMC) pp. 004716 - 004723
• Main Authors: Yamamoto, Yukiko, Itoh, Daichi, Tsuruta, Setsuo, Muranushi, Takayuki, Hada-Muranushi, Yuko, Kobashi, Syoji, Mizuno, Yoshiyuki, Knauf, Rainer
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2016
• Subjects: Case Base; Conferences; Cybernetics; Genetic algorithms; imbalanced data; Restart; Sociology; solar flare forecast; Statistics; Support Vector Machine; Support vector machines; Weather forecasting
• DOI: 10.1109/SMC.2016.7844976

Solar activity has various influences on the global environment. Specifically, it may have serious impacts on the Earth such as satellite damage, etc. and power plant failures causing more serious disaster. For a precise forecast of larger scale solar flares causing serious disaster, it is important to improve the space weather forecast, a daily forecast of the solar flare. In our work so far, a machine-learning algorithm called Support Vector Machine (SVM) was used. We extended this technology by integrating Case Based Genetic Algorithm (CBGA) for a more precise forecast. It was shown experimentally that triple mutation rate on the slowdown of evolution in our CBGA improves considerably (e.g. another 5%) more than original mutation rate in the True Skill Statistics TSS. For further obtaining the optimality towards more imbalanced data analysis applicable to the recognition of serious disaster or medical disease, Restart CBGA is proposed with its expected effect. Here GA integrating SVM is restarted using highly optimized but diversified solutions in the case base as initial individuals. Further this restart CBGA is repetitively and evolutionary performed, evolving and maintaining the case base by the result of each (restarted) GA.