Deep data: Key to driving big data success

Firstfuel, the CTO, says the search for smaller, more relevant condensed information is key to unlocking the vast potential of large data.

There is no doubt that everyone has heard of "big data", but "deep data"? The answer is probably no. Don't worry, I'm not going to give you a new professional vocabulary. But in view of the recent controversy over the amount of data that business users are continuing to collect and manage, I think the concept of deep data should go into the flat of your business users who are concerned about the potential of the data. Focus on the energy efficiency of the construction industry Analysis Enterprise Firstfuel company CTO and chief data Officer Badri Raghavan has its own unique insights. The company's clients, including government agencies and energy agencies, are using Firstfuel energy Analysis Services to promote greener and more cost-effective programs to the office environment, schools and other facilities.

In a telephone interview, Raghavan talked about his views on "deep data" and how Firstfuel companies turned the idea into a competitive advantage.

"What we call ' deep data ' is a complex of specialized knowledge reserves in the field--for us, the combination of the energy industry and the data science--to help technicians analyze the energy use of buildings from a macro-scale perspective," he told us.

The concept of deep data is closely related to the density of information. "There may be a lot of information in a given data stream," Raghavan said. "On the contrary, it is possible to collect a large amount of data that lacks sufficient conclusive content or information." ”

As you may have guessed, Raghavan himself does not agree with the http://www.aliyun.com/zixun/aggregation/12240.html "> Data collection or, as far as possible, the aggregation of more information." But at present many companies do so, that is, not sure whether the significance of the blind collection of large amounts of data.

The real core of data collection is efficiency, or "use of the data assets that are already available." To achieve this, we need to first identify which technologies or business challenges we need to address. What kind of data flow is most important among the resources that we can use?

A single data stream is often the most important measure of Firstfuel's industry-the analysis of the energy consumption of large buildings.

"We will use the measurement data as a scanning result for a building," he said. Using our data science algorithms, we can analyze the health status of buildings, identify the weak links and the areas where there is still room for efficiency improvement. ”

This, he notes, is an excellent example of how deep data actually works. Measured data is "a relatively concise flow of data, but it contains a wealth of content," Firstfuel was able to position its most interesting question: to identify energy consumption in violation of the efficiency of priority principles.

Of course, the most important thing for many businesses is to figure out which data streams are the most analytical, and then to combine them with other data to get a new analytical conclusion.

Firstfuel has found several types of data streams that are usually most potentially valuable.

"Metering data can tell us a lot about buildings," Raghavan said. "We're going to start using high-resolution aerial imagery--yes, that's Google Earth, where we use a lot of this stuff in our work." From our point of view, it contains a wealth of potential information. It tells us what types of equipment are on top of these buildings, "and Firstfuel is able to make a general judgment on the amount of energy that the building needs to consume."

The analyst also included data from the National Meteorological Service Center for consideration.

"We set about setting and gradually introduce it gradually." As long as the information analysis conclusions can be improved, we will take the relevant data flow into consideration. ”

That, according to him, is the basic concept of deep data. "You can do a deep research on a relatively small set of data, instead of having to face the vast sum of data as it has in the past ... And try to get a small needle that symbolizes a valuable conclusion. ”

For example, Firstfuel can collect a wide range of additional data-including information related to traffic flow and parking conditions, as well as Twitter data streams-but there is no clear reason to drive them in such laborious ways.

"A large data ocean that can potentially be analyzed in relation to direct jumps into the presence of massive amounts of data, often end up with little or no valuable information, we tend to gain more from relatively small amounts of data-focusing on data that actually reflects the objective state of the building," Raghavan points out. "After we have worked out such a solution, we will gradually turn our ideas into reality," he said. ”