Increasing the Reliability of Pattern Recognition by Analyzing the Distribution of Errors in Estimating the Measure of Proximity between Objects

Abstract

When recognizing similar or close objects in a report, the accuracy of the recognition is very low when the value of the measure of proximity between objects (MPBO) is close to the value of the error that occurs. modern algorithms are preferred instead of emprig dosturlar to improve accuracy in calculating the measure of proximity between objects. The algorithm proposed in previous research work is not effective, although it eliminates problems such as gross error, correlation coefficient, and the presence of a modular sign in formulas. Proposing a new methodology, range analysis was used instead of summarizing the results when calculating parameter values. The advantage of this system is distinguished by error reduction, more accurate recognition and efficiency. The given algorithm was modeled on a computer and the results were obtained. The processing of the results shows that, thanks to the proposed methodology, it is possible to significantly increase the accuracy of the calculation of the measure of the proximity between objects. At this time, it does not affect the running speed of the system.