Data Mining

WizSoft has developed a proprietary algorithm for revealing if-then rules (association rules) in a given data. The algorithm quickly reveals all the if-then rules that meet user-defined thresholds. An example of an if-then rule:

If City is NYC
and Amount Purchased is 200 … 300 (average=250)
and salesperson is Dave
Then
Growth Since Last Year is less than 0

Rule’s Probability: 0.70
The rule exists in 3,700 records
Significance Level: Error probability, 0.001

The last three numbers denotes the rule parameters:

The Rule’s Probability (sometimes known as the confidence level) designates the ratio between the number of records in which both the condition(s) and the result hold, and the corresponding number of records in which the condition(s) hold with or without the result.

The second number (sometimes called the support level) is simply the number of records where both the condition(s) and result hold.  

The third number Significance Level indicates the degree of the rule’s validity. It is equal to 1 minus the error probability which qualifies the probability that the rule exists accidentally in the data under analysis.

WizSoft also developed a propriety algorithm for revealing if-and-only-if rules (necessary and sufficient conditions). An example of an if-and-only-if rule:

The following conditions explain when Growth Since Last Year is less than 0

  1. Amount Purchased is 200 … 300 (average=250)
  2. Item is CAR235

If at least one of the conditions holds, then the probability that Growth Since Last Year is less than 0 is 90%

If all the conditions do not hold, then the probability that Growth Since Last Year is NOT less than 0 is 90% is 92%

Both rules are used for revealing interesting phenomena as well as issuing predictions for new cases.

Text mining

WizSoft has developed an algorithm for concept-based searching. This algorithm –

  • disambiguates words having several meanings
  • reveals the key words that summarize the contents in each section
  • finds sections that are relevant for a search query in natural language

WizSoft has developed an algorithm for finding words or names that are phonetically similar, which is stronger than the Soundex algorithm.

Bank reconciliations

WizSoft has developed an algorithm for revealing one-to-one, one-to-many, and many-to-many matches that meet user-selected criteria. This algorithm is used for issuing bank reconciliations and account reconciliations.

WizSoft holds thee registered patents in United States.

  1. PATTERN RECOGNITION USING GENERALIZED ASSOCIATION RULES Patent No: US 6,311,173 B1 Date of Patent Oct. 30, 2001.
  2. SYSTEM AND METHOD FOR REVEALING NECESSARY AND SUFFICIENT CONDITIONS FOR DATABASES ANALYSIS – Patent No: US 6,542,881 B1 Date of Patent:   Apr. 1, 2003.
  3. FINDING SUSPICIOUS ASSOCIATION RULES IN DATA RECORDS – Patent No. US 8,595,200 B2, Date of Patent  Nov. 26, 2013