turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Data Mining
- /
- Average Square Error is too high.

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

04-26-2014 03:20 PM

Hi,

I'm building a model which is giving me the very high average square error and misclassification rate.

Ho can I reduce these two results. Please provide me valuable inputs.

Also, Please let me know what is basic flow for Data Mining Model.

Thanks.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to NareshAbburi

04-26-2014 06:49 PM

That's way too difficult to answer here.

Your model is clearly not fitting well, so try changing the variables included in the model.

To learn more about data mining perhaps look into the CRISP-DM framework and/or check out the data mining courses offered on Coursera, EdX, Udacity for starters.

Lecture Notes | Data Mining | Sloan School of Management | MIT OpenCourseWare

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Reeza

04-26-2014 08:01 PM

Hi,

I've tried with multiple combinations, but still ASE is too high. it is nearly 2000. and my validation ASE also nearly 2000.

Is there any alternative to fit my model well.

Thanks.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to NareshAbburi

04-26-2014 09:29 PM

without any context its hard to say. How many variables do you have? What is your predictor? How many categorical variables are there? How many continuous? Have you standardized your variables? Or transformed them? What did the univariate analysis show? Are the scales of your variables incredibly different?

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Reeza

04-27-2014 05:21 PM

Hi,

I've a continuous target variable. Input variables are both continuous and categorical variables. (Continuous - 4, Categorical - 3)

I'm trying to build logistic regression. (Will it work...?)

I've standardized and applied transformation also to reduce skewness of the variables.

Plz suggest.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to NareshAbburi

04-27-2014 05:33 PM

No.

Logistic regression is for a binary target variable. Linear Regression is for a continuous target variable.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Reeza

04-27-2014 05:35 PM

Then what is best prediction model to apply for combination of categorical and continuous inputs..?

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to NareshAbburi

04-27-2014 05:39 PM

Linear regression. The model is more dependent on the output required than the input.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to NareshAbburi

04-27-2014 05:41 PM