Methods and apparatus to generate highly-interactive predictive models based on ensemble models.

A computing apparatus for generating predictive models regarding a product using Random Forest Modeling algorithms, with the apparatus allowing at least a user to enter the user's knowledge regarding the product, the apparatus comprising:

• at least one computing device;
• and
• one or more storage devices coupled to the at least one computing device, with the one or more storage devices storing instructions that, when executed, cause the computing apparatus to:
• receive source data regarding the product, with the source data categorized under a plurality of attributes;
• edit the source data;
• identify a factor to be predicted;
• select from the plurality of attributes at least some of the attributes as selected attributes based on the factor, with the list of selected attributes separated into logistic selected attributes and demographic selected attributes, and with the number of demographic selected attribute over the total number of selected attributes forming a percentage;
• generate a plurality of strong learning models via applying the Random Forest Modeling algorithms on at least a portion of the source data in the selected attributes;
• select one of the strong learning models as a first predictive model to predict the factor, with the first predictive model having a number of trees;
• update the source data based on results from a first marketing campaign after the launching of the first marketing campaign;
• randomly select from all the trees in the first predictive model to form a reduced number of trees, with the reduced number depending on the percentage;
• and
• update the first predictive model by applying the Random Forest Modeling algorithm to the selected trees based on at least a portion of the updated source data in the selected attributes,
• wherein the instructions when executed cause the computing apparatus to provide a plurality of user interfaces for the user to enter information at least regarding the product to assist in generating and updating the first predictive model.

A computing apparatus for generating predictive models to predict a factor regarding a product using Ensemble Modeling algorithms, with the apparatus allowing at least a user to enter the user's knowledge regarding the product, the apparatus comprising:

• at least one computing device;
• and
• one or more storage devices coupled to the at least one computing device, with the one or more storage devices storing instructions that, when executed, cause the computing apparatus to:
• receive source data regarding the product, with the source data categorized under a plurality of attributes, with the list of attributes separated into logistic attributes and demographic attributes, and with the number of demographic attributes over the total number of attributes forming a percentage;
• generate a plurality of weak learning models via applying the Ensemble Modeling algorithms on at least a portion of the source data for the factor, with each of the weak learning model depending on only a portion of the source data;
• generate a plurality of strong learning models, with each strong learning model including a different number of weak learning models;
• select one of the strong learning model as a first predictive model to predict the factor;
• update the source data;
• randomly select from all the weak learning models in the first predictive model to form a reduced number of weak learning models, with the reduced number depending on the percentage;
• and
• update the selected weak learning models in the first predictive model via applying the Ensemble Modeling algorithms on at least a portion of the updated source data for the factor;
• and
• generate a second predictive model to predict the factor by combining the updated weak learning models with the weak learning models not selected in the first predictive model,
• wherein the instructions when executed cause the computing apparatus to provide a plurality of user interfaces for the user to enter information at least regarding the product to assist in generating the first and the second predictive models.

A computing apparatus for generating predictive models regarding a product using the Random Forest Modeling algorithms, with the apparatus allowing at least a user to enter the user's knowledge regarding the product, the apparatus comprising:

• at least one computing device;
• and
• one or more storage devices coupled to the at least one computing device, with the one or more storage devices storing instructions that, when executed, cause the computing apparatus to:
• receive source data regarding the product, with the source data categorized under a plurality of attributes;
• edit the source data;
• identify a factor to be predicted;
• select from the plurality of attributes at least some of the attributes as selected attributes based on the factor, with the list of selected attributes separated into logistic selected attributes and demographic selected attributes, and with the number of demographic selected attribute over the total number of selected attributes forming a percentage;
• separate the source data under the selected attributes into a training set of source data and a validation set of source data;
• generate a plurality of trees via applying the Random Forest Modeling algorithms on the training set of source data for the factor, with each of the trees depending on only a portion of the training set of source data;
• generate a plurality of strong learning models, with each strong learning model including a different number of trees generated;
• select one of the strong learning models as a first predictive model to predict the factor, with the select depending on the validation set of source data;
• update the source data;
• randomly select from all the trees in the first predictive model to form a reduced number of trees, with the reduced number depending on the percentage;
• update each of the trees in the selected trees in the first predictive model via applying the Random Forest Modeling algorithms on at least a portion of the updated source data for the factor;
• and
• generate a second predictive model to predict the factor by combining all the updated trees with the trees not selected in the first predictive model,
• wherein the instructions when executed cause the computing apparatus to provide a plurality of user interfaces for the user to enter information regarding at least the product to assist in generating the first and the second predictive model.