The Jordan Pi-Sigma neural network for temperature prediction
In recent years, various temperature forecasting models have been proposed, which broadly can be classified into physically-based approaches and statistically-based approaches. Hitherto, those approaches involve sophisticated mathematical models to justify the use of empirical rules which make th...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2012
|
| Subjects: | |
| Online Access: | http://eprints.uthm.edu.my/2330/1/24p%20NOOR%20AIDA%20HUSAINI.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In recent years, various temperature forecasting models have been proposed, which
broadly can be classified into physically-based approaches and statistically-based
approaches. Hitherto, those approaches involve sophisticated mathematical models to
justify the use of empirical rules which make them less desirable for some
applications. Therefore, in this respect, Neural Networks (NN) have been
successfully applied and with no doubt, they provide the ability and potentials to
predict the temperature events. However, the ordinary NN adopts computationally
intensive training algorithms and can easily get trapped into local minima.
To overcome such drawbacks in ordinary NN, this research focuses on using a
Higher Order Neural Network (HONN). Pi-Sigma Neural Network (PSNN) which
lies within this area, is able to maintain the high learning capabilities of HONN.
The use of PSNN itself for temperature forecasting is preferably utilisable just yet.
Notwithstanding, this study disposed towards an idea to develop a new network
model called a Jordan Pi-Sigma Neural Network (JPSN) to overcome the drawbacks
of ordinary NN, whilst taking the advantages of PSNN. JPSN, a network model with
a single layer of tuneable weights with a recurrent term added in the network, is
trained using the standard backpropagation gradient descent algorithm. The network
was used to learn a set of historical temperature data of Batu Pahat region for
five years (2005-2009), obtained from Malaysian Meteorological Department
(MMD). JPSN’s ability to predict the future trends of temperature was tested and
compared to that of Multilayer Perceptron (MLP) and the standard PSNN.
Simulation results proved that JPSN’s forecast comparatively superior to MLP and
PSNN models, with lower prediction error, thus revealing a great potential for JPSN
as an alternative mechanism to both PSNN and ordinary NN in predicting
the temperature measurement for one-step-ahead. |
|---|