Salford Predictive Modeler

The SPM Salford Predictive Modeler software suite is a highly accurate and ultra-fast analytics and data mining platform for creating predictive, descriptive, and analytical models from databases of any size, complexity, or organization.

Languages supported:

10.0/10 (Expert Score) ★★★★★
Product is rated as #11 in category Predictive Analytics Software
Ease of use
Support
Ease of Setup

The SPM Salford Predictive Modeler software suite is a highly accurate and ultra-fast analytics and data mining platform for creating predictive, descriptive, and analytical models from databases of any size, complexity, or organization.

Salford Predictive Modeler
Salford Predictive Modeler

Show more categories

Customer Reviews

Salford Predictive Modeler Reviews

Bill K.

Advanced user of Salford Predictive Modeler
★★★★★
SPM--The best predictions you will ever make

What do you like best?

SPM is an exceptionally good tool for making predictions.

1) The user interface is both menu driven (for exploratory use) and also programatically driven (for highly efficient production.) The menu-system produces as output the corresponding program for saving, improvement, and reuse.

2) There is an exceptional data cleaning and exploratory toolkit (much better than standard tools like SAS's PROC FREQ). Well thought-out. Data cleaning is extremely important, even in data mining, to understand the many nuances, histories, and biases of the data.

3) To make economically-optimal decisions requires optimizing the error costs. You cannot simply minimize the total Type 1 and Type 2 error rates (or related functions like Somers' D or Gini) as the costs of the errors can be vastly different. SPM's CART is a fabulous program in that it enables this extremely important optimization.

4) The main tool of SPM is TreeNet--their Stochastic Gradient Boosting algorithm. TreeNet is genuinely exceptional. It allows efficiently exploring a rich family of algorithmic options. In particular it has a sophisticated way to explore the required interaction structure.

5) The Partial Dependency Plots are easy to produce and explore and enable developing an intuition for the general behavior of the predictions. You can then spline these plots (with monotonicity if so desired), put those new basis functions back into the fit, and end up discovering a set of simultaneous optimal transformations. This is not possible to do with any univariate tool.

6) The scoring code can be output in numerous languages and thus easily deployed in diverse production environments.

7) The additional data mining algorithms (Random Forests, Multiple Adaptive Regression Splines, and more) are well implemented and at times give additional views into the corners of your predictions.

SPM is absolutely the top-of-the-line of powerful, easy to use, flexible, reliable, data mining software. It is the standard that pretty much everyone is always comparing to and against.

I have brought it into multiple companies and everywhere it has been readily adopted by both the advanced predictive modelers as well as by the skilled business analysts.

What do you dislike?

I hope that the new owners of SPM (Minitab) continue to invest in the development of SPM. I have heard of some very nice new ideas that will make SPM still better.

What problems are you solving with the product? What benefits have you realized?

Most recently I have been been modeling consumer behavior in US banking. This involves responsible use of credit, retail price sensitivity, and responsiveness to direct and internet advertising. I have been using Bayesian Sequential Experimental Designs (along the lines of Thompson Sampling) to continually (re)evaluate the optimal set of treatments to be applied to every individual.

Review source: G2.com

Leave a reply

Your total score

B2B Software Guide