Big Data makes bi and predictive analytics more powerful

As we all know, predictive analysis has always been the "prerogative" of statisticians and data scientists in the "Ivory Tower", and they are far from day-to-day business decision makers. Big data will change that.

As more and more data streams are put online and integrated into existing BI, CRM, ERP, and other critical business systems, predictive analytics will eventually become a focus of attention. Although most customer service representatives and field sales representatives have not yet felt the impact, companies such as IBM and MicroStrategy have begun to move.

Big Data: Predictive analysis is no longer the prerogative of statisticians

Imagine a world where a customer service representative can independently determine whether a problem customer is worth keeping or upgrading, or a salesperson can tailor a retailer's product to the retailer's ratings on Facebook or Twitter.

Large data put more commonly used tools such as group analysis and regression analysis into the hands of day-to-day managers, and then they can use the non-transactional data to make strategic long-term business decisions.

Then, the big data is not meant to replace the traditional bi tool, said Rita Sallam, a bi analyst at Gartner Research, that big data will make bi more valuable and business-friendly, "We always need to look at past data, and when you have big data, you should do it." BI does not disappear, it is strengthened by large data. ”

How do you know that the predictions seen during the initial phase of the discovery will be proven over time, for example, in the Midwest, do red wallets really sell better than blue wallets? Preliminary data analysis may suggest that the red purse sells better because the red purse sold more in the previous quarter (or earlier).

But this is relevant and there is no causal relationship. If you look more closely at---Use the historical transaction data collected from the BI tool, you will find that it is actually the result of the latest business positioning activity because the business is looking at the red purse.

That's why IBM's emerging technology director David Barnes is more likely to refer to the results from large data technologies such as Hadoop, Map/reduce, and so on. For example, you don't want to make critical business decisions based on an emotional analysis of the Twitter stream.

Analyze unstructured data in social media to get direct returns

There are big business opportunities in social media. For example, as a retailer, you find Justin Bieber's analysis very much like the jacket he wore at last night's concert, and someone on Twitter said he bought one from your store, and then you can quickly decide to increase the amount of inventory on that coat, because you know this dress will be very popular, But only in a very limited time.

If there is no predictive analysis (PA), you are likely to miss this opportunity.

"In the past, we would make decisions based on historical data, but now the times are different," Barnes said. "Now we need predictive analysis." ”

We need to combine open source technology (where most large data platforms originate from open source), Moore's Law, commodity hardware, cloud computing, and the ability to capture and store large amounts of non transactional data.

Unstructured data, such as video and e-mail, which is often considered the driving force behind large data, is rarely involved in the process. You can brush blogs and user forums and then associate this information with geo-data and combine existing structured customer data with new sources such as the Micro Strategy Wisdom engine, which tracks 14 million of Facebook users ' evaluation of your brand. So you get a strong predictive power.

"There are two things that have happened to the big data," says P.k.paleru, marketing director at MicroStrategy, a bi supplier. "You may be able to combine various types of data from different sources, and you can also micro-tune all of these data." ”

Shorten the time of large data analysis

One of the great advantages of this analysis, says Paul Barth, founder and managing partner of New Vantage Partners, information management and analysis consultancy, is to shorten the "Answer Time" (TTA) Data scientists have taken months to build queries or models to answer forward-looking business questions about supply chains or production plans, and now it only takes a few hours to complete.

This is because large data technologies allow information to be analyzed before being optimized or relational. Coupled with advanced analytics, the business manager asks and answers questions in a very short time, but still needs IT staff and data modeler to lend a helping hand.

"These people are using big data to automate the process of machine learning," says Barth, which produces 20,000 data models per product line and market, allowing users to predict the next 18 months. "It's a big change and they're able to do this because large data technologies can automate many modeling steps and can be performed unattended." ”

Not long ago, it was almost impossible for a statistical analyst to build a single model for weeks or months. If you sell 100 products, your entire product line cannot exceed 1000 models, which means that the information returned by these models is not very accurate.

The golden age of Big data analysis has yet to come

While this is exciting for business users, large-scale data analysis techniques are not going to happen so quickly. Hadoop, though powerful, is still just the "raw" tool for processing massive datasets.

Thinking carefully about the usefulness of these predictive analysis results, does the opinion of 100 million people really exceed 100,000?

"There is a lot of duplication of data," says Barth, "If you want to analyze correctly," You still need smart analysts, and fortunately, big data provides them with very powerful tools. (邹铮 Compilation)

(Responsible editor: The good of the Legacy)