CGS-Mask: Making Time Series Predictions Intuitive for All
Authors: Feng Lu, Wei Li, Yifei Sun, Cheng Song, Yufei Ren, Albert Y. Zomaya
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluated CGS-Mask on synthetic and real-world datasets, and it outperformed state-of-the-art methods in elucidating the importance of features over time. According to our pilot user study via a questionnaire survey, CGS-Mask is the most effective approach in presenting easily understandable time series prediction results, enabling users to comprehend the decisionmaking process of AI models with ease. |
| Researcher Affiliation | Academia | 1School of Computer Science and Technology, Huazhong University of Science and Technology, China 2The Australia-China Joint Research Centre for Energy Informatics and Demand Response Technologies, Centre for Distributed and High Performance Computing, School of Computer Science, The University of Sydney, Australia 3Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, China |
| Pseudocode | Yes | Algorithm 1: CGS-Mask generation Input: The black-box prediction model f, two-dimensional cellular automata (CA) in size m n, N rounds of generations, the probability of crossover operator Pc, the probability of mutation operator Pm, and the probability of translation operator Pt. Output: The optimal mask M |
| Open Source Code | No | The paper mentions a 'Technical Appendix (Lu et al. 2023a)' for more details, but this reference points to an arXiv pre-print (arXiv:2312.09513), not an explicit code repository or a statement of code release. |
| Open Datasets | Yes | We conducted experiments using synthetic data sets, including the rare features and rare time data sets from (Ismail et al. 2020). ... The MIMIC-III data set (Johnson et al. 2016) contains the health record of 40,000 ICU patients at the Beth Israel Deaconess Medical Center. ... LSST simulated astronomical time-series data in preparation for observations from the Large Synoptic Survey Telescope (Emille et al. 2018). ... NATOPS data set is generated by sensors on the hands, elbows, wrists, and thumbs for gesture recognition (Ghouaiel, Marteau, and Dupont 2017). ... AE was obtained from the Appliances Energy Prediction data set from the UCI repository to predict the total energy usage of a house (Candanedo, Feldheim, and Deramaix 2017). |
| Dataset Splits | No | The paper uses various datasets and mentions that 'For a fair comparison, the data selection, preprocessing, and model training were the same as (Tonekaboni et al. 2020)'. However, it does not explicitly provide the specific percentages or sample counts for training, validation, or test splits within the paper itself. |
| Hardware Specification | Yes | We are also grateful to AMD product (China) Co., Ltd. and the Sugon Information Industry Co., Ltd. for their X785-g30 series GPU server. |
| Software Dependencies | No | The paper does not provide specific version numbers for software dependencies, libraries, or programming environments used to run the experiments. |
| Experiment Setup | Yes | In our experiments, we set D = T = 50. For the former two data sets |A| = 125, and for the latter two data sets |A| = 250. For simplicity, the perturbation value was set to zero. Since all tasks were regression, we evaluated the mask by Equation (4). ... For a fair comparison, the data selection, preprocessing, and model training were the same as (Tonekaboni et al. 2020). |