Application of artificial neural network for categorization of mobile bills
Session
Computer Science and Communication Engineering
Description
Businesses and organizations in general, are constantly trying to make improvements regarding effective costs management, by planning their costs efficiently. The aim of this study is to determine the utility of artificial neural network in making the categorization of mobile phone users based on their mobile bills. Data for the year 2015-2017 are transformed into two datasets; one dataset for training the model, and the other for testing the model. Neural network model is initiated, starting from less complex model, consisting of less layers and nodes, and is scaled-up on complexity by being tested continuously, until the best results are retrieved. On this basis, it is recommended that data from the datasets are as much evenly distributed as possible, when used for training the neural network model. The neural network model consists of 23 nodes on input layer, 230 nodes in first hidden layer, 23 nodes in second hidden layer, and 6 nodes in the output layer. After training with 500 epochs, the model produces these results: 81.3% for validation, 74.5% for testing, and 0.068% overfitting. Neural network model can be successfully implemented for categorization of mobile users based on mobile billing system. The results produced by the neural network model can be used for planning, reporting, and other organizational requirements.
Keywords:
neural network, machine learning, artificial intelligence
Session Chair
Bertan Karahoda
Session Co-Chair
Krenare Pireva
Proceedings Editor
Edmond Hajrizi
ISBN
978-9951-437-69-1
Location
Pristina, Kosovo
Start Date
27-10-2018 10:45 AM
End Date
27-10-2018 12:15 PM
DOI
10.33107/ubt-ic.2018.84
Recommended Citation
Gashi, Kastriot and Karahoda, Bertan, "Application of artificial neural network for categorization of mobile bills" (2018). UBT International Conference. 84.
https://knowledgecenter.ubt-uni.net/conference/2018/all-events/84
Application of artificial neural network for categorization of mobile bills
Pristina, Kosovo
Businesses and organizations in general, are constantly trying to make improvements regarding effective costs management, by planning their costs efficiently. The aim of this study is to determine the utility of artificial neural network in making the categorization of mobile phone users based on their mobile bills. Data for the year 2015-2017 are transformed into two datasets; one dataset for training the model, and the other for testing the model. Neural network model is initiated, starting from less complex model, consisting of less layers and nodes, and is scaled-up on complexity by being tested continuously, until the best results are retrieved. On this basis, it is recommended that data from the datasets are as much evenly distributed as possible, when used for training the neural network model. The neural network model consists of 23 nodes on input layer, 230 nodes in first hidden layer, 23 nodes in second hidden layer, and 6 nodes in the output layer. After training with 500 epochs, the model produces these results: 81.3% for validation, 74.5% for testing, and 0.068% overfitting. Neural network model can be successfully implemented for categorization of mobile users based on mobile billing system. The results produced by the neural network model can be used for planning, reporting, and other organizational requirements.