Inventory: the "God Prediction" and "Oolong prediction" in the history of large data

With the disease prediction, the World Cup forecast, "Big data" has become every it circle people can not leave the topic, as if everything towards the trend of Apple 6-"bigger than bigger", the bigger is the better. With the constant application of large data innovation, large data prediction is increasingly "myth", the 100% accuracy rate seems to be a hard indicator of the ability to measure large data, but that is not the case. Marvellous's big data predicts a moment of "epiphany" and a "oolong" moment. Today, the small series led us to take stock of the big data in the history of the critical moment.

"God" predictions in the history of large data

In 2009, Google launched the "Flu trend prediction" project, a few weeks before the swine flu outbreak, "Google flu trend" successfully predicted the spread of influenza in the United States, its analysis results even specific to specific regions and States, and very timely, so that public health officials alarmed. The move hit Google's position in big data forecasts.

3721.html ">2014 year, Baidu large data prediction team through the massive event information mining and analysis, the 2014 Brazil World Cup all 64 games of the outcome, as well as champions and Dark Horse forecasts. Whether the One-fourth finals or the 16 knockout, Baidu predicted the accuracy of the results reached 100%! Cut Goldman Sachs, win Google, the World Cup forecast thoroughly fire Baidu.

In terms of weather forecasts, the "prediction Emperor" is the Earthrisk company, which uses large data to forecast future weather conditions, using a predictive model different from the previous numerical forecast model, based on 82 billion calculations and 60 meteorological historical data to identify weather patterns, Then we compare these models with the current climatic conditions, and then use the predictive analysis to forecast the weather, which has longer prediction time, higher prediction degree, and can generate hot and cold weather probability 40 days in advance. The traditional subjective prediction model is earthrisk in front of the system.

The "Oolong" prediction in the history of data

From 2011 to 2013, Google's flu trend forecast began to decline, from August 2011 to September 2013 108 weeks, it overestimated the trend of influenza epidemic more than 50%. The "Google flu trend" did not work in the latest flu outbreak, the Nature magazine said. The tool has worked reliably for several winters, providing rapid and accurate outbreaks of influenza in the context of massive data analysis and so-called "no theoretical models". This time, Google's product display data, and the United States Centers for Disease Control after the summary of the results compared to exaggerate almost one times! The majority of projects overestimate the incidence of influenza-like diseases.

Hit by the trough of the flu forecast, Google forecasts again every enemy. In this year's World Cup forecast, before the One-fourth finals, Baidu, Google, Microsoft and Goldman Sachs on the World Cup of the four-strong forecasts, the results show that this time Baidu, Microsoft, Goldman Sachs three to accurately predict the results, and Google only accurately forecast the top four of the three teams. The prediction results are accurate, the role of the prediction model is very important, it is because Microsoft, Goldman Sachs, Baidu, Google and the use of different forecasting models, which led to the difference between their predictions. Compared with Baidu's high accuracy rate, Google, Microsoft, Goldman Sachs I am afraid to be together to reflect on this mistake.

However, the World Cup has just dispersed, Baidu at the end of September to launch the film box office forecast the first Test water also suffered a small embarrassment. In the product on-line stage, Baidu forecast "golden age" of the box office for 2.0~2.3 billion, October 16, "Golden Age" cumulative box office for 46.98 million (has been released 16 days), so the results for the film side, the media and the public are unexpected, for Baidu forecast is also a "oolong."

Today the Internet industry is surging, "Big data" hot. The premise of large data prediction is to admit its uncertainty, have right and wrong, success and failure. We cannot reject the technology and results of large data because of a deviation. While these projects are temporarily flawed, this is not enough to negate the potential contribution of large data to society. Any new things have a process of development. The value of large data prediction functions in all walks of life, some have been in tune with the right time and highlight, some still need to take time to achieve the ideal of people's requirements. We have reason to believe that the future value of large data will be the same as the trend of this era "bigger than bigger".