A non-transitory computer readable medium having instructions stored thereon that, when executed by a processor, cause the processor to perform customizable machine prediction using an extensible software tool, wherein, when executed, the instructions cause the processor to:
- receive a specification comprising at least features of a trained machine learning model including a set of input features used to generate predictions;
- generate an interface for the trained machine learning model based on the received specification, wherein the interface comprises one or more definitions that correspond to the set of input features;
- load the trained machine learning model using the generated interface, the trained machine learning model comprising a binary file configured to receive data as input and generate prediction data as output, wherein the loading includes validating that the binary file is configured to receive input that matches the set of input features;
- generate, by the software tool, predictions using observed data that is stored according to a hierarchical multidimensional data model by:
- inputting a) a portion of the observed data that matches the set of input features to the binary file to generate first data predictions, and b) a portion of the observed data to a forecast model to generate second data predictions, wherein,
- the trained machine learning model is trained using confidential data that is not accessible by the software tool and that is not used to train or configure the forecast model,
- the multidimensional data model comprises at least a measure dimension, product dimension, and one or more additional dimensions, and
- the first data predictions correspond to a first member of the measure dimension and the second data predictions correspond to a second member of the measure dimension, the first member and the second member of the measure dimension comprising different measures related to product demand;
- retrieve, by a software component configured via plan data, intersections of the hierarchical multidimensional data model, wherein the plan data defines that c) first of the retrieved intersections include the first data predictions and a first product or product line from the product dimension, and d) second of the retrieved intersections include the second data predictions and a second product or product line from the product dimension;
- display, in a user interface, e) the first retrieved intersections and the second retrieved intersections, and f) an aggregation data value that corresponds to another intersection of the hierarchical multidimensional data model, wherein the aggregation data value comprises an aggregation of one or more of the first data predictions and one or more of the second data predictions;
- generate an updated interface based on an updated specification that defines an other set of input features different from the set of input features;
- load, using the updated interface, an other version of the trained machine learning model comprising an other binary file, wherein the other version of the trained machine learning model is a retrained version of the trained machine learning model, and the loading includes validating that the other binary file is configured to receive input that matches the other set of input features;
- and
- generate, by the software tool, additional predictions by inputting, to the other binary file, an other portion of the observed data that matches the other set of input features.