Page 447 - Contributed Paper Session (CPS) - Volume 4
P. 447
CPS2564 Tiffany Rizkika et al.
3), natural cubic spline. Natural cubic spline is a continuous segmentation
function and remains continuous on all derivatives so that it makes it very
smooth.
Model 1 uses a parametric aprroach to detect outlier (IQR), so the resulting
model is called IQR.Spline. While Model 2 uses a non-parametric approach to
detect outlier (KDE), so the resulting model is called KDE.Spline. The time
series-based will use previous period data in modelling, using the Nowcast
Model proposed by Kim, Cha, and Lee (2017).
4. Result
a) Historical Data-Based (Volatile food price nowcasting)
1) Visualization Crowdsourcing Data
The picture below shows the number of crowdsourcing data report
in weekly periods every commodity. It shows that the number of report
is unstable that cause the frequency of data wasn’t same everytime. So
before modelling, data will be preprocessed.
2) Modelling with Training Data
Neural Network Neural Network Neural Network
DLM Lag 1
Commodity (2) (3) (3,2)
Daily Weekly Daily Weekly Daily Weekly Daily Weekly
Chicken 0,08332466 0,059171286 0,0827 0,0477 0,0831 0,0592 0,0834 0,0579
Beef 0,001968903 0,001844379 0,0020 0,0018 0,0020 0,0018 0,0020 0,0018
Egg 0,019875244 0,02287047 0,0171 0,0159 0,0171 0,0221 0,0174 0,0230
Chili 0,077096627 0,058217853 0,0578 0,1428 0,0757 0,1214 0,0651 0,1203
Onion 0,108276312 0,099677708 0,1024 0,0675 0,0970 0,0712 0,0937 0,0308
Low quality
rice 0,046325108 0,038070219 0,0266 0,0172 0,0260 0,0081 0,0256 0,0222
Premium
quality rice 0,060732601 0,058284652 0,0424 0,0233 0,0334 0,0165 0,0174 0,0147
436 | I S I W S C 2 0 1 9