11Ants Model Builder FAQ Support

How to Get Started

The first step in getting started is to read the Data Preparation Guidelines, then watch the tutorial videos. The Frequently Asked Questions (FAQ) section below contains answers to the most common questions.

Use the forum to communicate with the 11Ants Analytics community about how to use 11Ants Analytics products and learn about Machine Learning.

Still got a question then email 11Ants Analytics directly - click on the "Got a Question?" button above.

11Ants Analytics Forums

The 11Ants Analytics forums are there to help you make the most of 11Ants Analytics technologies.  You can discuss how to make the most of what 11Ants Analytics offers and mix with 'like minded' users and learn about predictive analytics with machine learning experts.  11Ants Analytics monitors the forums so please register and post your views and questions.

FAQ

Do I need Microsoft Excel to use 11Ants Model Builder?
Yes all user interface is via Microsoft Excel, so you will require Excel 2007 or Excel 2010.

To get a trial of Excel 2010 go to the Microsoft Office website http://office.microsoft.com

How do I Activate the 11Ants Model Builder License using my Key?

  1. Install 11Ants Model Builder
  2. Start up Microsoft Excel
  3. Select 11Ants Model Builder from the Excel menu
  4. From the 11Ants Model Builder ribbon menu select the Help --> License Activation...

  1. Now enter your validated e-mail address and the License Key that 11Ants Analytics e-mailed to your validated e-mail address.

  1. Press the 'Activate Now' button and 11Ants Model Builder will attempt to activate your copy. 

Note, you must be connected to the Internet for the activation process to complete successfully.

How do I get a License Activation Key?

Register for a trial or purchase a license from 11Ants Analytics.  Once registered with a validated e-mail address the License Activation Key will be e-mailed to your validated e-mail address.

Why is it called 11Ants Model Builder?
A model is a description of a complex entity - in this case, the complex entity is your data. 11Ants Model Builder uses our proprietary HyperLearn technology to search for subtle patterns in your data and then automatically generates thousands of models to accurately describe it, each of these models are evaluated and ranked. The top performing models effectively create order from chaos and enable insights previously unimaginable.

What types of data mining can I perform with 11Ants Model Builder?
11Ants Model Builder is particularly well suited to the following types of data mining:
Regression (predict a number), Classification (predict a category) and Rank Likelihood Analysis (predict and rank from most likely to least likely); as well as indicating which inputs are most useful in helping predict.

What is predictive analytics?
Predictive analytics is the branch of data mining concerned with the prediction of future probabilities and trends.

What is classification?
Classification is a data mining technique which develops a model that separates the data into the different classes, with the aim of classifying new examples in the future. For example a classification model can be built to classify flowers into various families based upon measurements of various parts of the plant.  You use classification data mining when the target column is categorical by nature.

What is rank likelihood analysis?
Sometimes we may be interested to rank items in order of most likely to least likely. For example, imagine that we are mining data with a view to developing a model which helps to predict which an insurance company providers are fraudulently billing them. By ranking likelihood we will line each provider up from most likely to be fraudulent, to least likely. This means that our audit team would have a clear intelligent work sequence to follow, that would likely yield the most productive results.

What is a predictive model?
A predictive model is a model which has been constructed with the express purpose of predicting values or classifications of unseen examples. For example a predictive model may be built to predict the fuel consumption of a vehicle based only upon inputs such as number of cylinders, engine size, vehicle weight etc. For an interesting example of this, see the Quick Start Tutorial Video.

What is Influence Importance?
You may be interested to know which of the inputs most impact the target value e.g. In the example above is it weight or number of cylinders which most impacts fuel consumption. 11Ants Model Builder calculates the relative influence of all variables and provides each of them with a percentage.

What is an ensemble?
An ensemble of models is best described as a committee of models. Rather than just having a single model, several models work together. The basic goal when designing an ensemble is the same as when establishing a committee of people: each member of the committee should be as competent as possible, but the members should be complementary to one another. 11Ants Model Builder's proprietary HyperLearn technology facilitates automated creation and evaluation of model ensembles.

What is a training set? What is a test set? And why do I need them?
During the course of data mining 11Ants Model Builder observes (or learns) patterns in the data. The structure of these patterns is then described as a model. It is fundamental rule of data mining that you should not test on data that was involved in building the model, as we are interested to see how well the data generalizes to external data. So data mining best practices dictate that you will split your data into a training set (the set you learn from), and a test set (an independent test set, that you can test models efficacy upon). 11Ants Model Builder allows one click partitioning of data into test set and training set - this is achieved in the PREPARE button, on the 11Ants Model Builder ribbon.

Why would I want to randomize my data?
Data mining, and predictive modeling in particular, relies upon a diverse training set. As an example a given data set may be sorted in order of least to highest. In that case if we partition data into a test set and a training set, the lower values may be in the training set, and the higher values may end up in the test set. We will be putting ourselves at an artificial handicap by applying a model that was built upon only low value items to a test set of predominantly high value examples. By randomizing the data we eliminate this concern and ensure the examples are evenly distributed. 11Ants Model Builder allows the user to automatically randomize data prior to partitioning by checking the "Randomize order of Rows" box behind the PREPARE button.

What is a prediction confidence?
Predictive modeling requires predictions to be made. By definition these predictions will be of varying accuracy. 11Ants Model Builder has a feature which allows a prediction confidence (high, medium, low) to be attached to every prediction. To enable this check the box "Compute confidence for each prediction" behind the PREDICT button.

When does model builder stop?
11Ants Model Builder will continue generating and evaluating models until you tell it to stop. In the Enterprise version there is functionality to trigger a final build and stop based upon a pre-defined threshold(s) being met.

What algorithms are used in model builder?
The following algorithms are used in 11Ants Model Builder:

Can I select the algorithms I wish to use?
All of the algorithms are evaluated automatically. Thee user can constrain which algorithms are used, simply by checking the boxes next to each algorithm they would like to be used. This enables users to focus on just a single algorithm if they wish to.

Can I see which algorithms were used to create the final model?
Yes

How is algorithm tuning handled?
Many of the algorithms used in this technology require tuning parameters to optimize the effectiveness of the algorithm. 11Ants Model Builder's proprietary HyperLearn technology automatically handles all algorithm tuning.

Is model builder restricted to any special type of data?
Generally no - 11Ants Model Builder is agnostic to data type.

How long does it take to build a model?
This varies dependent upon the size of the data set, the strength of the patterns (if any) found within the data, the level of precision required, and a number of other factors. Generally speaking on small data sets useful models can be obtained in minutes.

What does the quality indicator tell me?
This is provides a general guideline as to how well the model appears to be working. This is an estimate only, and is estimated using internal validation based solely upon the training set.

Does 11Ants Model Builder automatically know which inputs (features) are relevant?
11Ants Model Builder's proprietary HyperLearn technology enables 11Ants Model Builder to automatically evaluate which inputs are relevant and which are not. You can also ignore input columns if you know they are not relevant, by un-checking the box next to the appropriate input, behind the ANALYZE DATA & GENERATE MODELS button.

How do I make predictions?
Simply select the data you would like to predict from (assuming the columns are arranged in the same sequence as in your training set), press the PREDICT button, select the model you would like to use from drop down menu. Then press PREDICT NOW.

I would like to deploy a model developed on 11Ants Model Builder to a production environment, can I do this?
Yes - please contact 11Ants Analytics to discuss options, we have a number of solutions for this. Ranging from housing the model in our 11Ants Predictor software, to hard coding the resultant model so it can be embedded into a device.

Can I use Excel for other things while 11Ants Model Builder is running?
Absolutely - 11Ants Model Builder leaves Excel free to be used for other operations, even while generating models.

What is the maximum number of projects I can have open at one time?
Unlimited.

How does 11Ants Model Builder handle data that is to be used to produce a new prediction that is outside the boundary of the trained predicted mode?
The technical answer is... 'it depends'!
There is no real yes or no, like many things relating to data mining it is impacted by the characteristics of the specific data set, but the somewhat unchanging fundamentals are:

Why is your company called 11Ants Analytics?
Ants have an ability to self organize which is unparalleled in nature. Our proprietary HyperLearn technology is an incredibly complex self organizing methodology which almost miraculously creates order from chaos. Our proprietary HyperLearn technology comprises 11 base machine learning algorithms. And we liked the name!

I've tried data mining before with products like SAS and SPSS. Is 11Ants Model Builder as difficult to use?
No - absolutely not. These products were heralded as easy to use when they were first launched, and they really were relative to the alternatives. However with our proprietary HyperLearn technology we have added a supervisory layer and 11Ants Model Builder provides a whole new meaning to 'easy to use'. The first person we showed this technology to was somebody who had been selling data mining software for over 15 years, his comment: "I have never seen anything that is as simple to use" - that is the power of 11Ants Model Builder. Our promise is if you can mine use Excel you can mine your data!

How large data sets can 11Ants Model Builder handle?
Microsoft Excel has a limitation of 1,048,576 rows, and 16,384 columns. However for data sets larger than this you can usually extract a smaller sample to train from. This only refers to the model building process, predictions are a linear process so for deployment the system can be configured to handle tens of millions to hundreds of millions of predictions, using our 11Ants Predictor software.

Does 11Ants Model Builder work on Windows XP and Office 2007?
Yes it does.  However certain configurations it may not.  We recommend that that you download a trial of 11Ants Model Builder and try it out.  If it does not work try the following work around.

The solution is something that the end user has to apply.  This solution is related to Microsoft technologies and as such 11Ants Analytics cannot be responsible or guarantee the outcome.

      Note this is only for Windows XP running Microsoft Office 2007

  1. Install Windows XP service pack 3
  2. Install office 2007 service pack 1 (sp1) and service pack 2 (sp2)
    1. sp1: http://www.microsoft.com/downloads/details.aspx?FamilyId=9EC51594-992C-4165-A997-25DA01F388F5&displaylang=en
    2. sp2: http://www.microsoft.com/downloads/details.aspx?FamilyId=B444BF18-79EA-46C6-8A81-9DB49B4AB6E5&displaylang=en
  3. Modify the register key as described in article "A 2007 Office system application does not load an add-in that is developed by using VSTO" (http://support.microsoft.com/kb/976811/)
  4. Install the hot fix (976477) "Description of the 2007 Office system hot fix package: October 27, 2009" (http://support.microsoft.com/kb/976477/)

 


© 2010 11Ants Analytics Limited.  All Rights Reserved