Does big data cloud computing make predictions more accurate?

Human life needs to be predicted, but reliability is really too much to be flattering and rarely true. There are human factors and technical reasons.

such as "non-sample error." Suppose a driver, driving the age of 30, travel 20,000 times, only occurred 2 times slight rubbing accidents. The Mid-Autumn Festival with the family drink a lot of wine, then this driver can because of previous driving record is good, think this will not accident? Clearly this is the wrong idea. Since the 20,000-time travel record is no alcohol driving record, this time the drink is too much, the previous record has no statistical significance. It may be that such a low-level error prediction expert can avoid it, but it is not. The 2008 global financial crisis caused by the United States, only one or two people predicted, and all the other U.S. rating agencies, White House think-tank, economists can not predict. The reason is that this "non-sample" prediction error. When the situation changes, blindly based on past records to make predictions, you can only get the wrong answer.

Many people like to invest in the stock market. In a bull market, investors may also be able to earn some money, but from the bull market into a bear market, securities companies are generally a collective mistake. This is more of a human factor. Securities analysts are wrong to judge the normal, but make mistakes must avoid only their own mistakes, together to make mistakes is tantamount to not making mistakes. For example, some people analyze the stock market has a certain probability to collapse, the best strategy is to continue to hold. So the stock market crashes, because most of the peers are unsure when to crash, also choose to hold strategy, collective mistakes, and will not show their low level. But if you sell the stock rashly, the short-term share price does not fall or even rise, it can only show that they are not enough.

Shocked by the global "http://www.aliyun.com/zixun/aggregation/23688.html" > terrorist attacks, it is very sudden, in fact, the United States intelligence agencies almost see through this major conspiracy. August 16, 2001-Mousavi, a religious extremist, was arrested. He only conducted 50 hours of flight training, but asked to participate in the Boeing 747 aircraft simulation training. It was so bizarre that people reported it. In hindsight, the signal was clear and terrorists were going to blow up the building by plane. At the time, the signal was masked in hundreds of thousands of of such noises, not prominent, maybe he was just a flying enthusiast. The signal, the more noise, makes the prediction very difficult.

All of these factors lead to inaccurate predictions, but there are ways to make predictions come closer to the truth by using the Bayes theorem. The probability theorem has been produced for more than 200 years, and it is to use conditional probability inference problem to reveal people's cognitive process and law of probabilistic information, and to instruct people to make effective study and judgment decision. For example, a female breast X-ray shows positive, so what is the probability of her breast cancer? Existing statistics show that if a woman does not have breast cancer, the probability of X-ray positive is 10%, if you do have breast cancer, X-ray positive probability is 75%; But if the Bayesian theorem is used to analyze, she is only 10% of the probability of breast cancer, because more than 40-year-old women, breast cancer probability is very low, only 1.4%, that is, the priori probability is very low.

Large data age, although the information explosion, but the signal and noise coexist, to make the right prediction is not easier than before, or even more difficult. "Signal and Noise," a book tells us, if the Bayesian theorem as the basis for efforts to understand the cause and effect of things, to avoid some of the wrong people or technical errors, the prediction accuracy rate will improve a lot.