Feature Selection Optimization Solar Insolation Prediction Using Artificial Neural Network: Perspective Bangladesh
Journal Title: American journal of Engineering Research - Year 2016, Vol 5, Issue 8
Abstract
This paper presents a new climatological feature selection model using artificial neural network by utilizing the real world data for Bangladesh, called as FSOSIP. In this analysis back propagation algorithm is applied. To facilitate the search strategy FSOSIP model uses 12 different combinations of 10 climatological features and a series of experiments were done to select the best subset of relevant features. The monthly averaged data has been collected from Bangladesh Metrological Department for seven stations as Dhaka, Barisal, Chittagong, Khulna, Rajshahi, Rangpur and Sylhet. The 1176 data between 2000-2013 are used to train the neural networks while the 365 data from 2014 are used to test the neural network. Experimental results show that the proposed FSOSIP model can select only two salient features easily with increasing the prediction accuracy which are relative humidity and maximum temperature. Furthermore, to prove the robustness of the model seven different models for seven stations of BMD are used and the results are then being averaged. Minimum MSE (0.000173%) and MAPE (0.0868%) shows the higher efficiency to predict the solar insolation. In addition, the proposed model exhibit better performance with feature selection strategy than any other model used up to now in Bangladesh.
Authors and Affiliations
Md. Shahid Iqbal1 ,, Md. Ataullah Mishkat2 ,, S. M. Rashedul Islam3 ,, Md. Shariful Islam4
Keywords
Related Articles
- EP ID EP403959
- DOI -
- Views 94
- Downloads 0
Segmentation Brain Tumor and Diagnosing Using Watershed Algorithm
Brain cancer is the most life impacting and most critical type of malignancy considering the sensitivity of the location in which it affects, the brain being the most important part of the body yet its cells lack the abi...
Performance Evaluation of Some Noise Prediction Modelling Approaches: Multiple Linear Regressions and a Hybrid Approach
Noise prediction models are very useful for urban planning and environmental management. As a result, researchers are always searching for methods that are practically applicable in predicting noise levels accurately. It...
Structural Investigation into the Causes of Cracks in Building and Solutions: A Case Study
Cracks are commonly found on and in structures and are usually undesirable features naturally occurring due to age, workmanship and other natural and environmental causes. This study is set to characterize cracks with th...
Health of the spine: a review about our basis
The foundation of the human body, the spine presents several characteristics that imply in everyday and sports practice. This structure, besides having ligaments and muscles, is also part of the nervous system, due to th...
About Tenth Problem of D. Hilbert
In this article the author prove unsolvability of tenth problem of Hilbert