training tesseract ocr 4.1.0

[jasw...@sukshi.com]

hi, 

im currently using tesseract version 4 .1.0 to train models. I observed that each time i train my model with same data and keeping same parameters also..everytime i end up with different character error rate and different accuracy. I want to get the best of my data..can anyone help please?