Pengembangan Model Protis Neural Network Untuk Prediksi Dan Klasifikasi Data Timeseries Dan Image

Abstract

Currently, the most widely developed methods for prediction and classification are Deep Learning, Machine learning and Artificial Neural Networks. The development of Neural Networks is currently growing very rapidly. Currently, Neural Network has many methods such as Convotional Neural Network, Reccurent Neural Network, Radial basis function and others. The problem that often occurs is that the results are variable and unstable, both in terms of performance and accuracy depending on the modified model, data, and variables used. Furthermore, it takes a long time to determine the level of layers to be used, the number of filters used, and so on. So that an in-depth analysis of the current methods is needed to compare them with several methods that have been used and try to do the best analysis that can be used, especially for time series data and images that will be used for classification and prediction. The analytical method used is to introduce a method derived from the workings of the protists and adapted to the Artificial Neural Network. Furthermore, the method will be tested into several data, namely time series and images to analyze the quality of the developed model. The analysis is expected to be an input for discoveries in improving the prediction and classification performance of the existing system, especially for time series data and image data. The results of the implementation of the model with the Cifar dataset show an accuracy rate of 0.695 or 69.5%, the results in the figure are still not as good as the results of implementing the model on time series data, namely the Precision value of 0.952 and the recall value of 0.950. So a more in-depth analysis is needed but the value has exceeded 50%. A more in- depth analysis of the development model is needed