COMPARISON OF TREND PROJECTION METHODS AND BACKPROPAGATION PROJECTIONS METHODS TREND IN PREDICTING THE NUMBER OF VICTIMS DIED IN TRAFFIC ACCIDENT IN TIMOR TENGAH REGENCY, NUSA TENGGARA

Aleksius Madu

Abstract


The purpose of this study is to predict the number of traffic accident victims who died in Timor Tengah Regency with Trend Projection method and Backpropagation method, and compare the two methods based on the degree of guilt and predict the number traffic accident victims in the Timor Tengah Regency for the coming year. This research was conducted in Timor Tengah Regency where data used in this study was obtained from Police Unit in Timor Tengah Regency. The data is on the number of traffic accidents in Timor Tengah Regency from 2000 – 2013, which is obtained by a quantitative analysis with Trend Projection and Backpropagation method. The results of the data analysis predicting the number of traffic accidents victims using Trend Projection method obtained the best model which is the quadratic trend model with equation Yk = 39.786 + (3.297) X + (0.13) X2. Whereas by using back propagation method, it is obtained the optimum network that consists of 2 inputs, 3 hidden screens, and 1 output. Based on the error rates obtained, Back propagation method is better than the Trend Projection method which means that the predicting accuracy with Back propagation method is the best method to predict the number of traffic accidents victims in Timor Tengah Regency. Thus obtained predicting the numbers of traffic accident victims for the next 5 years (Years 2014-2018) respectively - are 106 person, 115 person, 115 person, 119 person and 120 person.

 

Keywords: Trend Projection, Back propagation, Predicting.


Full Text:

PDF


DOI: https://doi.org/10.26486/mercumatika.v1i1.188

Article Metrics

Abstract view : 340 times
PDF - 547 times

Refbacks

  • There are currently no refbacks.




Jurnal Mercumatika : Jurnal Penelitian Matematika dan Pendidikan Matematika Indexing by

 ipiii.pngDimensions

 

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.