Next-generation data exploration tool based on LLM and the PCA/PLS family
Drag in an Excel file and the AI automatically identifies the title row, name row, observation columns, and X/Y variables, then creates the instance, configures the model, runs the fit, and lays out the charts in one go — your first model, with zero configuration.
Built-in large language model for one-click interpretation of complex statistical indicators such as T² and SPE, deep analysis of abnormal business root causes, and automatic generation and retention of structured analysis reports, with Markdown / Word export.
Describe your industry, plant, and process, and the AI directly produces a scenario priority scoring table and a data collection template; you can also upload a configuration diagram screenshot to automatically recognize the parameters in it and generate a data template.
The lasso tool visually selects outliers in charts, removes them with one click, and recalculates the model in real time; it also supports AI automatic outlier selection and comparative analysis of the red and blue groups.
Full coverage of PCA, PLS, PLS-DA, OPLS, and OPLS-DA; built-in S-Plot marker screening, SHAP interpretability analysis, and DML double machine learning causal inference.
Model exploration supports MV/DV variable marking, boundary constraint validation, optimization path replay, and 3D response surfaces, directly back-calculating "to reach the target quality, what should the process parameters be set to".
Copy a row or a batch of data from Excel and press Ctrl+V to paste it straight into the prediction panel for results; prediction results can be pasted back into Excel with one click, with no need to repeatedly export files.
Standalone closed-loop operation, with core algorithms and data processing completed locally to eliminate data upload to the cloud; already adapted to Xinchuang (domestic innovation) environments such as Kylin ARM, and models support custom packaging and download.