Optimasi Evolving Connectionist System dalam Multivariate Adaptive Regression Splines untuk Prediksi Perilaku Industri Financial Technology

Abstract

Evolving Connectionist System (ECoS) is a model of Evolving Intelligence Systems (EIS) which has an adaptive and continuous learning method by developing its own structure and function in a machine learning structure. Multivariate Adaptive Regression Splines (MARS) is a model that has been developed for several regulations that produce a functional basis and can be said to be an optimal predictive model and often uses business problem data. Where, Financial Technology, which is often referred to as FinTech, is included in big data problems in the business sector, which can be modeled on the predictions of users and the FinTech industry. FinTech with various criteria that can be utilized, namely P2P Lending. So that the trend of FinTech in its use can be predicted by the industry using the MARS approach. However, the MARS results, which are a functional basis for predicting the behaviour of the FinTech industry, have been successful and obtained accuracy with an MSE of 0.020 and a GCV of 0.058. After obtaining the results of the functional basis, it is followed by a machine learning algorithm that is based on the ECoS principle and has been successfully carried out with various combinations of parameters where the maximum accuracy is in the learning rate 1 parameter with 0.3, learning rate 2 with 0.6, sensitivity threshold with 0.3 and error threshold with 0.1. of 0.433175207. and seen in the MARS model which is then followed by ECoS, the actual value is not much different from the predicted value. So that this algorithm is able to predict the behaviour of the FinTech industry.