If the data sample follows a "power law distribution", we can Bermuda Email List change the data distribution to normal by applying the "logarithmic transformation" technique. Since most of the features in our dataset have a "right-skewed" distribution, we only need to use the function as below logUse the Numpy library to change the distribution to normal form: ## log transformation - power law distribution .
you can see in the image below, the distribution of 10 examples of dataset features has been normalized to an acceptable level: Histogram chart of features after applying the logarithmic transformation technique Histogram plot of features after applying the logarithmic transformation technique - "click image to enlarge".