search

boathit / deepgtt

DeepGTT: Learning Travel Time Distributions with Deep Generative Model
36 stars 22 forks source link

The performance of DeepGTT #6

Open learnurban opened 3 years ago

learnurban commented 3 years ago

I have tested your code in the Didi of Chengdu dataset. However, the performance (MAE-6.12) is worse than DeepTTE(MAE-1.68). However, in your paper, you introduce your results better than DeepTTE. I check the code several times and don't find any errors in preprocessing and model. Can you provide the experimental code for Chengdu DIDI data?

learnurban commented 3 years ago

If you have time, please contact me. I can provide the entire code of my test for Chengdu DiDi data.

learnurban commented 3 years ago

Thank you for your reply. Maybe you can find the key of my worse results!

boathit commented 3 years ago

I have never conducted experiments on the Chengdu dataset and we only used the Harbin dataset in our paper. But as far as I know, the GPS points of the Chengdu dataset adopt a different coordinate system rather than the standard WGS-84, did you transform the coordinates before doing map matching?

learnurban commented 3 years ago

Yes, I transformed the Gaode GPS coordinate into standard WGS-84. I have tried many parameters of the xtsep and ystep. But the results of deepGTT are still poor. Also, I mean that your results of Deep GTT are better than DeepTTE in your paper, not Chengdu DIdi dataset.

learnurban commented 3 years ago

In other words, in your Harbin dataset, you mentioned that DeepTTE has poor performance. Do you know what causes DeepTTE to perform poorly in the Harbin dataset?

boathit commented 3 years ago

I will try to test the models on the Chengdu dataset and figure out why DeepGTT fails.

learnurban commented 3 years ago

Thank you very much! I look forward to your reply!