Please, be more specific. Explain what you mean by more number of data class? However, it is important to note that a larger sample size does not guarantee high goodness of fit. Also, creating more data class creates more variables. if this is the case, a statistician can detect this and may view this as some kind of econometric manipulation. However, a rule of thumb could be to create the number of data class you want to, and then follow a stepwise regression by including and omitting some variables simultaneously to observe the value of R. Your goal should be to obtain a high goodness of fit with fewer variables. This is more considered to be scientific. 

I hope this helps? 

if i wish to test the accuracy of chi square test with more number of classes , will it improve the accruacy?

It would help if you would describe the variable you are using and the kind of analysis you are doing.