Dear Friends,
This is my first post of 2009... happy new year!
I have been absent this month because I’m writing the second chapter of my master's thesis. Anyway, today I’ll write a bit of neural networks, which are on of the most important and useful Data Mining techniques.But what is a neural network? A Neural Network (NN) is a system composed of many simple processing elements operating in parallel which can acquire, store, and utilize experiential knowledge. As I told you before, I’m very fascinated with Data Mining and the intelligence we can get from it. You can use it in Financial Markets (is the theme of my master dissertation), on insurance, or medicine, on retail, on all activities!!! The topics below explain better where you can apply NN, the learning methods, types and an Arquitecture that a NN could has.
NN Applications
-Classification
-Clustering
-Function approximation
-Prediction
NN Type
- Static (feed forward)
- Dynamic / Back-Propagation (feedback)
NN Arquitecture
- Single layer
- Multilayer
- Recurrent
- Self-organized
NN Learning Methods
- Supervised (network has inputs and their corresponding outputs)
- Unsupervised
In Data Mining classes directed by Professor Dr. Duarte Trigueiros, I did a practical excel exercise (homework) in order to better understand NN functionalities. I’ll share it with you today and feel free to contact me to get a copy of excel file. The excel files has the formulas behind and could be interesting if you think to improve your knowledge on this.
Initially, weights are random and the output of neural network will be compared to desired output. If same, reinforce patterns and if different, NN will adjust weights.
The dataset should be divided for training and testing, which in this case, because the NN is a supervised with back-propagation, the training dataset has the output classified and a better model will be created. After the model is created, the test dataset is considered as the input and the model could be evaluated.
Regards,
Pedro
