A Data Science Central Community
With the current interest in "Big Data" it is easy for the public to get the impression that "more data means greater value", which can be interpreted as better ROI, greater efficiency, and accurate decisions coupled with immediate positive ACTIONS.
But is this always the case?
The following discussion dissects this into 4 major points that have to be taken into account in order to get VALUE out of DATA !!!!! - WORTH TAKING A LOOK AT -
The example used to illustrate this is: ".....using predictive analytics in healthcare to mediate hospital readmissions......"
I should have added an interesting sentence that is in the cited link:
"But a well-known Chinese proverb states, “jùng dōu shòu dōu, jùng gwà, shòu gwà.” (If you plant beans, you get beans. If you plant squash, you get squash.) Algorithm and computer types know this better as “garbage in – garbage out.”
So True !!!!!
This articles makes a lot of good points. I have seen this time and time again, in which the targeted analysis makes much more sense then the large scale generic analysis. That is because domain expertise comes into play, and larger amounts of data become distracting and at odds with the expertise of the researcher. For important critical healthcare work, it becomes important that You control the data that you analyze, not the other way around.
Also, Insight IS important, however the problem with Big Data currently is that there is now way to quickly verify the validity of any insights provided. Sometimes if you cross reference the result with another Big Data insight, you will get contradictory results.
More abstractly, and less healthcare specific, the value of a 'big data approach' is that the performance differences between different ML algorithms become attenuated when the training set grows. The second learning from big data experiments is that the best algorithms are composites of many different algorithms.
The simple guidance, which big data experiments have provided, is that you need to start with the business decision that you want to resolve, then articulate the analytics that will enable you to answer the question, and that will drive your data collection and data transformation technology selection. Many big data efforts that turn this upside down fail.
Thanks for sharing the link. It has really broaden the perspective how data is valuable when it comes to examine it for predicting the trends. I would also like to share practical point of view of the same topic. For more information you can visit the below url:-