Effect of Training functions of Artificial Neural Networks (ANN) on Time series Forecasting

Abstract

Weather forecasting has been an area of considerable interest among researchers since long. A scientific approach to weather forecasting is highly dependent upon how well the atmosphere and its interactions with the various aspects of the earth surface is understood. Applicability of artificial neural networks (ANNs) in forecasting has led to tremendous surge in dealing with uncertainties. This paper focuses on analysis and selection of various techniques used in developing a suitable feed forward neural network for forecasting 24hr ahead hourly temperature using MATLAB 7.6.0 neural network toolbox. The data of 60 days hourly data temperature is used to train and test the different models, training functions, activation functions, learning functions, performance functions and the most suitable combination is selected. The performance and reliability of these models are then evaluated by number of statistical measures. Results are compared with each training function.

Comparing artificial neural network algorithms for prediction of higher heating value for different types of biomass

“…The comparative analysis of different algorithms for training an ANN aimed for time series forecasting implemented in the MathWorks MATLAB neural network toolbox, performed in (Aggarwal et al 2015) resulted with the conclusion that among all factors that affect the ANN's performance (amount of input data, the network complexity, activation function, number of hidden layers, number of weights and biases, etc), the training function is the most important factor for the accuracy of the network. Similarly, for predicting HHV of the biomass, all training functions exhibited reasonable accuracy, while the best performance has the one based on the Levenberg-Marquardt algorithm.…”

Computer vision for purity, phenol, and pH detection of Luwak Coffee Green Bean

“…For this reason, a training error of the learning function is carried out. The research results of Sharma and Venugopulan [39] and Aggarwal and Rajendra [40] 3 presents trainlm as the selected learning function which produces the lowest MSE validation. Trainlm is a learning function that updates the weights and biases based on Lavenberg Marquadt optimization.…”

“…The output layer expresses the percentage of purity, total phenol and pH in Luwak coffee. Designing the best ANN topology is accomplished through sensitivity analysis with a variety of learning functions; activation function; learning rate and momentum (0.1, 0.5, 0.9); hidden layer (1, 2); hidden layer node (10,20,30,40) with the lowest validation of MSE parameter. This study formulates the 3 activation functions i.e.…”

Analysis of Cities Data Using Principal Component Inputs in an Artificial Neural Network

“…This study considers data of 26 cities where 14 Travel parameters and 9 Land-use parameters are gathered as in Table 1 and Table 2 The methodology of principal components has been taken as established in (Paul et al, 2013) and (Manage and Scariano, 2013). Artificial neural networks and training algorithm are well described in (Pham and Sagiroglu, 2001), (Sharma and Venugopalan, 2014) and (Aggarwal and Kumar, 2015). (Burns and Whitesides, 1993) explained the concept of patterns and application of feedforward neural network.…”