Coupling Implicit and Explicit Knowledge for Customer Volume Prediction

Authors: Jingyuan Wang, Yating Lin, Junjie Wu, Zhong Wang, Zhang Xiong

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The effectiveness of GR-NMF in coupling all-round knowledge is verified over a real-life outpatient dataset under different scenarios. GR-NMF shows particularly evident advantages to all baselines in location selection with the cold-start challenge. Extensive experiments are conducted on a real-life outpatient dataset obtained from the Shenzhen city of China. The results show that GR-NMF outperforms competitive baselines consistently in various application scenarios with different sampling rates.
Researcher Affiliation Academia Jingyuan Wang, Yating Lin, Junjie Wu, * Zhong Wang, Zhang Xiong School of Computer Science and Engineering, Beihang University, Beijing, China School of Economics and Management, Beihang University, Beijing, China Research Institute of Beihang University in Shenzhen, Shenzhen, China Email: {jywang, linyating, wujj, wangzhong, xiongz}@buaa.edu.cn, *Corresponding author
Pseudocode No The paper describes the inference method using mathematical equations and textual descriptions, but it does not include any structured pseudocode or algorithm blocks with explicit labels like 'Pseudocode' or 'Algorithm'.
Open Source Code No The paper does not provide any explicit statement about making its source code open, nor does it include a link to a code repository.
Open Datasets No The paper states: 'We perform our experiments on an outpatient service data set collected from the public hospital system of Shenzhen, a major city in southern China1.' The footnote 1 links to Wikipedia page for Shenzhen, not the dataset itself. No specific link, DOI, or formal citation is provided for public access to this collected dataset.
Dataset Splits No The paper describes its experimental setup where 'sampled elements' (yij=1) are known and 'unsampled elements' (yij=0) are treated as unknown for prediction and evaluation. It mentions varying 'sampling rate from 10% to 50%', which implicitly defines known vs. unknown data. However, it does not explicitly state conventional training, validation, and test splits with specific percentages or sample counts for reproducibility. No distinct validation set is mentioned.
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., CPU, GPU, memory, cloud platform) used to conduct the experiments.
Software Dependencies No The paper does not specify any software dependencies with version numbers, such as programming languages, libraries, or specific solvers used for implementation or experimentation.
Experiment Setup Yes To set a proper H, therefore, we take a warmup experiment on a sampled dataset with 10% of all hospitals, and watch the predictive precision of GR-NMF with H varying from 5 to 40. As can be seen from Fig. 1 , when H > 20 the increasing trend of the predictive precision of GR-NMF tends to be flattened. As a result, we set H = 20 as the default setting in the following experiments. The objective function of GR-NMF is min J = Y (X S C) 2 F + α Y (X A k w) 2 F + β Y (A k w S C) 2 F + γ w 2 2 + δ S 1 + ζ C 1 s.t. S 0, C 0, w 0, where α = σ2 X1 σ2 X2 , β = σ2 X1 σ2 W 2 , γ = σ2 X1 σ2 W 1 , δ = σ2 X1 σ2 S , ζ = σ2 X1 σ2 R , which can be well estimated in advance by minimizing J1, J2 and J3 separately. In other words, these parameters are to be set before the optimization of J .