If the productized AI tool is ultimately GUI-style presentation, its main purpose is Visual Analytics, because visualization is the way computers convert data into information, transmit it and make it understandable. Applying visual analysis to data or models can achieve explainable, explorable, learnable, and participatory effects:

Interpretable: Data or models can use visualization to self-explain, discover patterns or explain anomalies; limitations of visualization Because there are few dimensions in a single analysis, multiple visualizations are usually displayed together;

Explorable: on the basis of visualization, unexpected surprises can be found through global display and interactive operations such as dimension combination, custom sorting, etc.;

Learnable: visualization is low-dimensional, so it can lower the threshold and improve learning efficiency for novices to understand data/models;

Participatory: models trained on past data cannot meet the prediction of new data, and the same applies to different scenarios Also, keeping the model accurate in different scenarios is a long-term process, and user participation can help with training.

The embedded PDF below corresponds to specific cases.