(Q) What is a model?
A model is any analytical routine designed to solve a specific business problem. Models in data mining typically involve supervised modes (regression, classification, neural networks,…) or unsupervised models (clustering, similarity reasoning, association rules,…) that can be created using a variety of tools (e.g. Python, R, Excel, SAS, etc.).
(Q) What are some typical use cases for predictive models?
Use-cases for predictive models are abound:
Predicting response rates to support targeted campaigns and optimize marketing spent
Personalizing special offers
Making product recommendations
Identifying cross or up sell opportunities
Evaluating credit risk
Sorting or ranking sales prospects
Forecasting sales or support requests
Optimizing social media strategies
(Q) What are the system requirement for “DM analytics”
Windows 64 bit
At least 3 GB RAM (memory requirements may depend on the size of the analysis file)
(Q) How my input data should look like?
For modeling process you need to create a “Flat file”, sort of a matrix, with each row corresponding to the modeling entity (customers, web browsers, loan takers,…) and each column representing an attribute In the world of Big Data, this matrix may contain millions, if not more, rows, and hundreds, even thousands, of attributes.
Check our Blog “Customer life Time value” for a real life example
(Q) I’m getting error message when loading a delimited file
Check that you set the right delimiter before selecting the “Next” button
(Q) Why I cannot move from the “Meta data” step to the “Model Definition” step?
To get an access to the “Model Definition” you must set a “Key” and “Target” in the “Meta Data” screen
(Q) Which model I should select in the “Model Definition” step?
For continuous target, like revenue, select the “linear” for True/False target select the “logistic”
(Q) Why I cannot select “Validation” and “Scoring”?
“Validation” and “Scoring” are available only after model has been built.
(Q) Why “Score New Data” is not working on my input file?
Your “Score data” file should contain exactly the same attributes as in the training dataset used to build the model.