product
Application for evaluation machine
H2O.ai
H2O dot AI
Driverless AI function introduction (demo)
Driverless AI Feature List
Automatic visualization of data
For the uploaded dataset, data relationships etc. are automatically visualized. Therefore, you can intuitively understand the data and get an overview of the data before creating a model. Outliers, histograms, correlation graphs, heatmaps, etc.
Automatic feature design
It automates the improvement work (feature design) of the dataset itself, which is important for improving the accuracy of the model. This task of designing new features using various techniques from the originally given data set largely depends on the skills, experience, and ideas of the data scientist, but Driverless AI automates this part. and does not require advanced user skills.
Model judgment reason visualization
When applying the created model in a business scene, isn't there a need to explain the prediction results of the model? To prevent AI from becoming a black Box, Driverless AI allows you to visually interpret the parameters that influence the prediction results.
◎ K-LIME, Shapley, Variable Importance, Decision Tree, Partial Dependence models can be explained as a whole, as well as for each cluster and individual results.
Model results reporting
It automatically generates the reports necessary to explain the created model to the business side and management. The details of the data used to create the model, the algorithm, and the model tuning flow are explained. It eliminates the need to create time-consuming validation reports, so you can move your model to production in less time.
In particular, regarding feature value design, the importance of feature values, the feature value design techniques used, etc. will be explained.
Execution module generation for Python and Java environments
Driverless AI automatically generates an execution module for the Python and Java environments for implementing the created model in existing internal systems and new systems.
- Python : Python Scoring Pipeline
- Java : MOJO Scoring Pipeline
Supports time series data
Models can also be created for time-series data, where temporal relationships between data are important. For time-series datasets, Driverless AI automatically detects the time order.
If necessary, the user can arbitrarily change the following settings.
- *In the case of sales forecast
- Group by columns: I want to forecast sales for each store and department
- How many weeks to predict: I want to predict sales after XX weeks
- When should we start predicting : 学習用データと検証用データとの時間差を〇〇週間つけたい
