Predictive Analysis Newbie

[ahsankhalid]

Hi Alteryx, Community, 

I am fairly new to alteryx  and have been using it for basic report creation and  pre  defined mathematical calculations for finance reports. 

I am moving to the predictive side of Alteryx and one of my projects is to predict if a lead will turn into an opportunity or not. 

We have 8 years of old data which I can use as a model data. 

Can anyone point me towards some relevant threads on this or similar kind of topics.

Kind Reagards

Ahsan Khalid

[HossC · Accepted solution]

Hey @ahsankhalid,

Here are a few resources that @SydneyF and I have pieced together:

First, we recommend checking out the One Stop Shop for Predictive Resources, which has links to a ton of awesome content involving our predictive tools!

Second, for your specific situation, it sounds like you are trying to predict a binary variable (a categorical variable that can take on 2 values: yes the lead turned into an opportunity or no the lead did not turn into an opportunity). Here is a useful thread on developing a binary classification model.

There are several modeling techniques available in Alteryx for estimating a binary variable, including: Logistic Regression, Boosted Model, Decision Tree, Forest Model, Naive Bayes Classifier, Neural Network, Spline Model, and Support Vector Machine. Alteryx has a built in workflow that demonstrates all of these models, which can be found in designer by following this path: Help -> Sample Workflow -> Predictive Analytics -> 11 New Donor. However, since you are new to the predictive tools in Alteryx, we suggest that you focus on a single tool at first.

If you are not interested in the variable selection process, the Forest Model may be the best tool to start with (this model can automate variable selection). Alteryx has two built in workflows for the Forest Model (Help -> Sample Workflow -> Predictive Analytics -> 7 Forest Models and Help -> Sample Workflow -> Predictive Analytics -> 12 New Donor Score Sample), and a handy explanation of the output can be found here. Here is a video that walks through the process of building a random forest model: Build a Random Forest Model.

However, if you are interested in specific variables and/or don't mind manually choosing variables, Logistic Regression may be the best tool to begin with. A discussion on logistic regression can be found here, and a useful suggestion in that thread is to use the predictive starter kit, which can be downloaded here (Starter Kit). Here is a video that walks through the process of building a logistic regression model: Logistic Regression Using Alteryx.

As a final thought - keep posting to Community with questions you have along the way! It is a really powerful resource, and definitely here to help you! :)

Thanks and welcome to the exciting world of predictive!

-Hoss Carroll and Sydney Firmin