Forum Stats

  • 3,733,004 Users
  • 2,246,671 Discussions
  • 7,856,461 Comments

Discussions

How to deploy Data Mining Project

usman_noshahi
usman_noshahi Member Posts: 163
edited May 2014 in Machine Learning

Hi,

I am new in DataMining. I have completed a tutorial on good customers if they buy new offers or not from a video tutorial but it's very similar to this

http://www.oracle.com/webfolder/technetwork/tutorials/obe/db/12c/r1/dm/ODM12c-PQ.html?cid=8307&ssid=122587621791540

Prediction(classification). I understood it completely. after connecting classification models to output view or table I have desire results. and I had table named Prediction_view in my schema

I wanna ask some questions.

  1. How can I apply these predictions on newly inserted customers?
  2. Every time I need predictions, should I work on project again and make fresh applied view or table?
  3. After using deploy (right click on data output view or table) I had all .sql files in my directory. Are those file used for fresh prediction algorithms?
  4. is there any way to refresh that view before making prediction?

Actually I have same scenario in my application. customers and sales quotation management. after sending sales quotation to customer I wanna apply Classification weather customer will respond to that quotation or not ?

thanks.

Best Answer

  • Denny Wong-Oracle
    Denny Wong-Oracle Member Posts: 183
    Accepted Answer

    Please refer to this white paper about how to deploy workflow using generated sql scripts

    http://www.oracle.com/technetwork/database/options/advanced-analytics/odmrcodegenwhitepaper-2042206.pdf

    http://www.oracle.com/technetwork/database/options/advanced-analytics/companion-2042207.zip

    I will use the workflow in the white paper as an example to answer your questions.

    1. How can I apply these predictions on newly inserted customers?

    In the demo workflow, the INSUR_CUST_LTV_SAMPLE APPLY node contains new data to be applied (Apply node) to the model (Class Build node) for predictions.  In your case, you may have a Data Source node that contains newly inserted customer data.

    2. Every time I need predictions, should I work on project again and make fresh applied view or table?

    In the demo workflow, the SCORED_CUSTOMERS node captures the predictions.  Notice the node creates the output as a VIEW (SCORED_CUSTOMERS_V), this view always reflect the predictions for new data.  There is no need to run any of these nodes again for new data, the view SCORED_CUSTOMERS_V always captures the predictions for latest data.  Had the node created the output as TABLE instead, you would need to run the node to persist new predictions.

    3. After using deploy (right click on data output view or table) I had all .sql files in my directory. Are those file used for fresh prediction algorithms?

    Please refer to the white paper for the script files usage.

    4. is there any way to refresh that view before making prediction?

    See above answers.

    Thanks,

    Denny

Answers

  • Denny Wong-Oracle
    Denny Wong-Oracle Member Posts: 183
    Accepted Answer

    Please refer to this white paper about how to deploy workflow using generated sql scripts

    http://www.oracle.com/technetwork/database/options/advanced-analytics/odmrcodegenwhitepaper-2042206.pdf

    http://www.oracle.com/technetwork/database/options/advanced-analytics/companion-2042207.zip

    I will use the workflow in the white paper as an example to answer your questions.

    1. How can I apply these predictions on newly inserted customers?

    In the demo workflow, the INSUR_CUST_LTV_SAMPLE APPLY node contains new data to be applied (Apply node) to the model (Class Build node) for predictions.  In your case, you may have a Data Source node that contains newly inserted customer data.

    2. Every time I need predictions, should I work on project again and make fresh applied view or table?

    In the demo workflow, the SCORED_CUSTOMERS node captures the predictions.  Notice the node creates the output as a VIEW (SCORED_CUSTOMERS_V), this view always reflect the predictions for new data.  There is no need to run any of these nodes again for new data, the view SCORED_CUSTOMERS_V always captures the predictions for latest data.  Had the node created the output as TABLE instead, you would need to run the node to persist new predictions.

    3. After using deploy (right click on data output view or table) I had all .sql files in my directory. Are those file used for fresh prediction algorithms?

    Please refer to the white paper for the script files usage.

    4. is there any way to refresh that view before making prediction?

    See above answers.

    Thanks,

    Denny

  • usman_noshahi
    usman_noshahi Member Posts: 163

    thanks for you reply. Really helpful for me. thanks.

This discussion has been closed.