This research proposed a method of selecting an optimal combination of numbers of input (number of lagged values in the model) and of hidden nodes for modeling seasonal data using Artificial Neural Networks (ANN). Three data sets (rainfall, relative humidity and solar radiation) were used in assessing the proposed procedure and the resulting ANN models were compared with two traditional models (Holt-Winter’s and SARIMA). Models with large number of lagged values have shown tendency to outperform those with small number of lagged values. Selected ANN model was found to outperform the two traditional models on rainfall data; it performed better than SARIMA but worse than Holt-Winter’s model on relative humidity data and performed worse than the two methods on solar radiation data. The proposed procedure has hence, performed fairly well. Oscillatory performance recorded by ANN models that resulted from the proposed procedure in relation to the other two models only attests to the fact that no particular model is best on every data set. Rather than insist on elegance or sophistication, researchers should be guided by parsimony.
Coronavirus outbreaks affect human beings as a whole and can be a cause of serious illness and death. Machine learning (ML) models are the most significant function in disease prediction, such as the Covid-19 pandemic, in high-performance forecasting and used to help decision-makers understand future situations. ML algorithms have been used for a long time in many application areas that include recognition and prioritization for certain treatments. Too many ML furcating models are used to deal with problems. In this study, predict a pandemic outbreak using the ML forecasting models. The models are designed to predict Covid-19, depending on the number of confirmed cases, recovered cases and death cases, based on the available dataset. Support Vector Machine (SVM) and Polynomial Regression (PR) models were used for this study to predict Covid-19 ‘s aggressive risk. All three cases, such as confirmed, recovered and death, models predict death in Ethiopia over the next 30 days. The experimental result showed that SVM is doing better than PR to predict the Covid-19 pandemic. According to this report, the pandemic in Ethiopia increased by half between the mid of July 2020. Then Ethiopia will face a number of hospital shortages, and quarantine place.
Mutual Fund Forecast Analysis. Case of Indonesia ETF (Exchange Traded Fund) Period February 2014 – July 2017 (Published)
Performance of mutual fund could be reflected from several factor, one of which are groth of Net Asset Value (NAV) and price of said mutual fund. The NAV growth of Indonesia Exchange Traded Fund (ETF) are considered one of the most significant, reaching 286% growth value in 2016. Besides, ETF industry in Indonesia itself are still at initial stage of growth. Therefore, information regarding this investment instrument are fairly unknown to public. This research aim to analyse price forecasting of Indonesia ETF using ARIMA method. Weekly price data collected and analysed within period of February 2014 to July 2017, and then forecasted for the next ten weeks. Sample of this ETF are taken from actively traded mutual fund in Indonesia Stock Exchange. Results shows that forecasting model of ARIMA (0,1,1) are the best model to forecast Premier ETF price on ETF IDX30, Sharia Premier ETF JII, and ETF Premier LQ45. Meanwhile, ETF Premier Indonesia Consumer, the best forecasting model are by using ARIMA MA(1) MA(4). This results are also supported by accuracy analysis, which shows that each of ETF Premier reached over 80% of accuracy.
Application of Neural Networks in Weather Forecasting (Published)
Weather Forecasting is the task of determining future state of the atmosphere. Accurate weather forecasting is very important because agricultural and industrial sector largely depend on it. Weather forecasting has become an important field of research in the last few decades. In most of the cases the researcher had attempted to establish a linear relationship between the input weather data and the corresponding target data. The Neural Networks package supports different types of training or learning algorithms. In this paper, the application of neural networks to study the design of neural network technique for Kanyakumary District,Tamil Nadu, India. A total of ten years of data collected for training the net work. The network is trained using the Back propagation Algorithm, Radial Basis Function, Regression Neural Network, Optical Neural Network, and Fuzzy ARTMAP Neural Network. The Fuzzy ARTMAP network can give the best overall results in terms of accuracy and training time. It is better correlated compared to the BPN,RBFN,GRNN and ONN networks. The proposed Fuzzy ARTMAP neural network can also overcome several limitations such as a highly non-linear weight update and the slow convergence rate.
Modeling and Forecasting of Armed Robbery Cases in Nigeria using Auto Regressive Integrated Moving Average (ARIMA) Models. (Published)
We have utilized a twenty-nine year crime data in Nigeria pertaining to Armed Robbery, the study proposes crime modeling and forecasting using Autoregressive Integrated Moving Average Models, the best model were selected based on the minimum Akaike information criteria (AIC), Bayesian information criteria(BIC), and Hannan-Quinn criteria (HQC) values and was used to make forecast. Forecasted values suggest that Armed Robbery would slightly be on the increase
This study modeled the inflation rates in Nigeria using Box Jenkins’ time series approach. The data used for the work ware yearly collected data between 1961 and 2013. The empirical study revealed that the most adequate model for the inflation rates is ARIMA (0, 0, 1). The fitted Model was used to forecast the Nigerian inflation rates for a period of 12 years. Based on these results, we recommend effective fiscal policies aimed at monitoring Nigeria’s inflationary trend to avoid damaging consequences on the economy.
Detection of Faulty Sensors in Wireless Sensor Networks and Avoiding the Path Failure Nodes (Review Completed - Accepted)
For variety of applications, Wireless Sensor Networks (WSNs) have become a new information collection and a monitoring solution. Faults occurring due to sensor nodes are common due low-cost sensors used in WSNs, deployed in large quantities and prone to failure. The goal of this paper is to detect faulty sensors in WSNs and avoiding the path failure nodes. Fault detection is based on the local pair-wise verification between the sensors monitoring the same physical system. Specifically, a linear relationship is shown between the output of any pair of sensors, when the input of a system comes from a common source. Using this relationship, faulty sensors may be detected by using forecasting model based on the parameter (i.e., temperature) and it also identifies which sensor is normal or abnormal. After the fault nodes are detected, first of all disable all the faulty nodes so that network is not affected by erroneous reading and send the information to the base station. Due to the nature of proposed algorithm, it can be scaled to large sensor networks and also saves energy from reduced wireless communication compared to the centralized approaches
Forecasting a financial time series, such as stock market trends, would be a very important step when developing investment portfolios. This step is very challenging due to complexity and presence of a multitude of factors that may affect the value of certain securities. In this research paper, we have proved by contradiction that the Nigerian stock market is not efficient but chaotic. Two years representative stock prices of some banks stocks were analyzed using a feed forward neural network with back-propagation in Matlab 7.0. The simulation results and price forecasts show that it is possible to consistently earn good returns on investment on the Nigerian stock market using private information from an artificial neural network indicator.