Being able to predict future trends is undoubtedly a useful tool to have. Being able to know an outcome ahead of time sounds great, but conducting an effective prediction is complicated. Here are some of the limitations of using prediction in Big Data:
Reliant on Accurate Data
Predictive algorithms are highly reliant on accurate, testable data. Veracity and variability are essential when conducting prediction. If the results are skewed even slightly, it can offset the entire prediction. This means the outcome can change drastically if the anomalies aren't picked up on. Furthermore, data that fluctuates may also change the outcome, as some data algorithms aren't advanced enough to analyse complex shifts/spikes in patterns.
Reliant on Historical Data
Another limitation would be the dependence on previous data. This might not be a problem for everyone, but could be problematic for businesses. Without any previous data, predictions cannot be made which means launching new stores or products cannot benefit from prediction. The same can be said with any application without former data, a foundation must be laid first.
Consequences of Inaccurate Prediction
Inaccurate prediction can result in many issues, and in some cases it may even cause harm. This is not cool. Areas of application include healthcare, where assumptions and shortcuts cannot afford to be wrong. This can cause physical damages or misdiagnoses to patients. It can also pose a threat in science, which has little room for error. Slight mistakes in prediction can have adverse and unforeseen results which costs a lot of time and money. Lastly, it can pose a threat to people financially. Investors rely heavily on prediction and means that if a predictive algorithm fails, it puts their capital and assets at risk.
References:
https://ilearn.fife.ac.uk/course/view.php?id=9751
No comments:
Post a Comment