Python Data Analysis Cookbook

Streaming counting with the Count-min sketch

Streaming or online algorithms are useful as they don't require as much memory and processing power as other algorithms. This chapter has a recipe involving the calculation of statistical moments online (refer to Calculating the mean, variance, skewness, and kurtosis on the fly).

Also, in the Clustering streaming data with Spark recipe of Chapter 5, Web Mining, Databases, and Big Data, I covered another streaming algorithm.

Streaming algorithms are often approximate for fundamental reasons or because of roundoff errors. You should, therefore, try to use other algorithms if possible. Of course in many situations approximate results are good enough. For instance, it doesn't matter whether a user has 500 ...