Kurtosis is another characteristic that defines how your data distribution looks like. It is the characteristic of being flat and spread out or being slim and peaked.
It is a measure of whether data is heavy-tailed or light-tailed in a normal distribution. So, if Kurtosis is high, it means that the tails of the distribution is getting more extreme values than the tails of a normal distribution. Hence the height of the distribution will be short and it will more spread, leading to a high standard deviation of maybe 6 or 7. 
Similarly, if the Kurtosis value is very low, the tail of the distribution will be less lengthier than the tail of a normal distribution (less than 3 standard deviation), resulting in a slim and taller height of the distribution.
A large value of Kurtosis is often considered as more risky as it might take outliers inside rather than disregarding them. Therefore, the data might tend to give an outlier value as outcome with greater distance from the mean if applied to any machine learning algorithm. So look at your data closely!
#statistics #datascience #machinelearning