ICACT20260040 Question.1
Questioner: yennt@hvnh.edu.vn 2026-02-09 오후 6:07:13 |
ICACT20260040 Answer.1
Answer by Auhor cytsai@saturn.yzu.edu.tw 2026-02-09 오후 6:07:13
Chrome  Click!! |
| The paper evaluates performance using HR, MRR, and NDCG. Could you please elaborate on how the ground truth for the Next-POI recommendation task is defined, and how the data are split into training and testing sets to prevent temporal information leakage? |
In most cases, many prepossessing procedures might be applied to the dataset before building the model. Thus, there is no really ground truth for my case. However, we can still compare the performance of different models to know how well my proposed method work.
|
ICACT20260040 Question.2
Questioner: ygt991129@mail.ustc.edu.cn 2026-02-09 오후 6:13:15 |
ICACT20260040 Answer.2
Answer by Auhor cytsai@saturn.yzu.edu.tw 2026-02-09 오후 6:13:15
Chrome Click!! |
| How can the generalization ability of this POI recommendation model be evaluated across datasets with diverse geographic and cultural backgrounds to ensure its robustness in practical applications? |
The proposed POI recommendation model is evaluated based on popular dataset from a social media called Yelp. Although devise geographic and culture background might be different and might have different impact to the result, the performance of the proposed model should be similar. |
ICACT20260040 Question.3
Questioner: ygt991129@mail.ustc.edu.cn 2026-02-09 오후 6:18:47 |
ICACT20260040 Answer.3
Answer by Auhor cytsai@saturn.yzu.edu.tw 2026-02-09 오후 6:18:47
Chrome Click!! |
| How can multimodal data such as user reviews or textual descriptions be integrated to enhance the semantic accuracy and interpretability of POI recommendations? |
The multimodal data such as user reviews or textual descriptions can be considered as a textual embedding for the POI. Therefore, we can use pre-trained language models such as BERT to extract the textual embbedding from reviews or descriptions. Then we can concate it with other embeddings to form more advanced feature representations. |
ICACT20260040 Question.4
Questioner: ygt991129@mail.ustc.edu.cn 2026-02-09 오후 6:22:25 |
ICACT20260040 Answer.4
Answer by Auhor cytsai@saturn.yzu.edu.tw 2026-02-09 오후 6:22:25
Chrome Click!! |
| How can the impact of special events like holidays on user POI selection behavior be modeled to dynamically optimize temporal features? |
When we construct temporal embedding, we might add "holiday" and "non-holiday" to the one-hot encoding. Although I did not try it, I believe it might workable. |
ICACT20260040 Question.5
Questioner: arjunvirvashisht@gmail.com 2026-02-10 오전 12:48:46 |
ICACT20260040 Answer.5
Answer by Auhor cytsai@saturn.yzu.edu.tw 2026-02-10 오전 12:48:46
Chrome Click!! |
| Which embedding type (temporal or geographic) had the most significant impact on your NDCG results? |
|
