In data dictionary, we release our data used in this paper. The release dataset collection and processing process can refer:

@misc{liu2020datadriven,
    title={A Data-driven Approach for Constructing Multilayer Network-based Service Ecosystem Models},
    author={Mingyi Liu and Zhiying Tu and Xiaofei Xu and Zhongjie Wang},
    year={2020},
    eprint={2004.10383},
    archivePrefix={arXiv},
    primaryClass={cs.CY}
}

Related code be found in:

• MSEM-EventExtraction
• MSEM-EvolutionaryRelationGenerate

data/nodes.json: Each line represents a node in the json format. Each node contains following fields:

data/edges.json: Each line represents an edge in the json format. Each edge contains following fields:

Service ecosystem snapshots generation, static community detection, nodes' social position calculation are time consuming. So we provide a copy of our intermediate results, which have been stored in data/*.pkl. You can use this data to save your time.

• generate_snapshots.py: this script mainly contains aging function and Service Ecosystem snapshots steps
• community_features.py: this script contains some common used community features.
• community_options.py: this script contains community detection, meta community identifier, community evolutionary event detection, community features vector construction, etc.
• model_operations.py: this script contains how to train the service community evolutionary event prediction model. And apply the SHAP explainer on the model.
• report.py: this script mainly contains how to visualize the analysis results.
• test.py: this script gives an example of whole process. (Notebook version will come soon)

Generated figures are stored in figure folder with .html format.

• summary_{evolution_evnt}.html: the summary report of evolution_evnt shows the importance of different features (merged) to the event.
• dependency_{feature}_{evolution_evnt}.html: the dependency report gives much detailed information.

Please refer to the paper to get more details.

In our paper, we chose random forest. You can chose other prediction mode. All you need to do is moddify the function train_prediction_model in model_operations.py.

if model_name == "random_forest":
    model = RandomForestClassifier(n_estimators=50)
elif model_name == "xxxx": #modify here, and change params value
    model = XxxxClassifier(mmmm=xxx)