Monthly Topic


predisoft has developed a plug-in module for Microsoft’s SQL Server Analysis Services environment. The plug-in adds classical “PCA” (Principal Components Analysis) functionality to Microsoft’s own suite in a seamless fashion.

For more information please watch our small video here.

Articles


Data Types

Other Developments


oroMATH has a great number of programs for data analysis and descriptive knowledge mining. More recent developments are oriented towards the implementation of predictive data analysis. These programs will be very useful to local and international companies to predict demand, inventories and risks in loan-granting.



Loan-granting scoring


A loan scoring program is being developed based on the correspondence factorial analysis and the discriminant analysis. Proposed by Gilbert Saporta, a French statistics-mathematician, uses the correspondence factorial analysis to transform qualitative variables (from a questionnaire filled in by loan applicants) into numeric variables used as the starting point in a discriminant analysis to discover the new variables that will discriminate between good and bad payers. These new variables are the factorial axes of the discriminant analysis and correspond to a linear combination of the original questionnaire variables, but at the same time different from all of them.



Demand Analysis through Time Series


Many types of data in the business world and in economy correspond to observations made during fixed intervals of time. A group of this type of data is known as a time series. For example, the Dow-Jones industrial average and the daily load in an ATM are time series. Business predictions based on time series are one of the most difficult problems faced by data analysis.


oroMath has methods that apply state-of-the-art statistical and mathematical methods to make prediction based on time series by implementing tendency study methods.


Personal Profiles Based on symbolic databases, oroMATH has implemented programs to summarize related databases. These methods allow realtime detection of client atypical behavior because the comparison of a new client transaction with a symbolic profile (a summary of his/her history) is extremely fast.