AI SearchApplication of back propagation method in forecasting problems
Jong Jek Siang
Application of back propagation method in forecasting problems - Bangkok : Asian Institute of Technology, 1992 - 65 leaves - Thesis ; no. CS-92-9 . - Asian Institute of Technology. Thesis ; no. CS-92-9 .
A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science, School of Engineering and Technology
Thesis (M.Sc.) - Asian Institute of Technology, 1992
In this study, monthly water quality (Temperature, pH, Conductivity) at Vientiane, Laos, and monthly water flows at Vientiane, Ubon, Yasothon, Wat-Tai Kosum and Tha Sang Kran Bridge are forecast one month ahead, using Back Propagation method without other external data. It was found that for seasonal data like water flows, small network (one hidden layer with one to three units in it) is enough to learn the pattern data. Additional hidden units do not improve the performance significantly. The results show also that generally, forecasting using Back Propagation method gives better results than using the average data in each month, and using data in the same month of the previous year. In addition, the result of Vientiane station shows also that forecasting using Back Propagation model gives a better results as compared to the Box-Jenkins method.
Neural networks (Computer science)
Forecasting
Application of back propagation method in forecasting problems - Bangkok : Asian Institute of Technology, 1992 - 65 leaves - Thesis ; no. CS-92-9 . - Asian Institute of Technology. Thesis ; no. CS-92-9 .
A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science, School of Engineering and Technology
Thesis (M.Sc.) - Asian Institute of Technology, 1992
In this study, monthly water quality (Temperature, pH, Conductivity) at Vientiane, Laos, and monthly water flows at Vientiane, Ubon, Yasothon, Wat-Tai Kosum and Tha Sang Kran Bridge are forecast one month ahead, using Back Propagation method without other external data. It was found that for seasonal data like water flows, small network (one hidden layer with one to three units in it) is enough to learn the pattern data. Additional hidden units do not improve the performance significantly. The results show also that generally, forecasting using Back Propagation method gives better results than using the average data in each month, and using data in the same month of the previous year. In addition, the result of Vientiane station shows also that forecasting using Back Propagation model gives a better results as compared to the Box-Jenkins method.
Neural networks (Computer science)
Forecasting