Large data forecast 4 characteristics, 10 typical industries

During the World Cup, the world's major technology giants used big data to predict the World Cup results, really let big data in the World Cup thoroughly fire. Don't think the World Cup is over, the big data forecasts will not stop. From predicting all kinds of phenomena to night-time astrology, from weather forecasts to plane crashes, crystal balls from the beautiful fairy-tale world to the present technology to predict the future. As the information revolution continues to deepen, predictions in the big data age have become simpler, and human life has entered the era of big data predictions.

Prediction is the core value of large data

When people talk about the collection, storage and mining of large data, the most common application cases are "predicting stock market", "predicting influenza", "predicting consumer behavior", and predictive analysis is the core function of large data.

Large data also has the function of data visualization and large data mining, mining the information value that has occurred and assisting decision-making. The traditional data analysis and mining are doing similar things, but the efficiency is lower or the depth, breadth and precision of mining are not enough. Large data forecasts are based on large data and predictive models to predict the probability of a future event. Turning the analysis from "Facing the past" to "facing the future" is the biggest difference between big data and traditional data analysis.

The logical basis for large data predictions is that every unconventional change must have a sign beforehand, that everything is traceable, and that if the law of the signs and changes is found, predictions can be made. Big Data predictions are not sure that something is necessarily going to happen, it is more of a probability.

Four conditions for large data forecasts viewed from the weather forecast

There are predictive analyses based on big data before the Internet: Weather forecasts. Because of the Internet, the following several characteristics of the large data forecasts represented by the weather forecast are embodied in more fields.

1. The timeliness of large data prediction. Weather forecast granularity from day to hour, there are strict aging requirements, based on massive data through the traditional way of calculation, come to the conclusion that tomorrow has come, the forecast is not worth. Large data prediction applications in other areas have higher requirements for "timeliness", such as equities and real-time pricing, while cloud computing, distributed computing and the development of supercomputers provide such high-speed computing power.

2. Data source for large data prediction. Weather forecasts need to collect a large amount of meteorological data, meteorological satellites, weather stations are responsible for collection, but the whole system of deployment and operation of a huge cost. There are few areas before the Internet that have such data collection capabilities. WEB1.0-centric information generation, WEB2.0 for social creation, mobile Internet is anytime, anywhere, socialization and multi-device data upload, the cost of each evolution data collection is greatly reduced, the scope and scale is greatly enlarged. When large data is detonated, the data source required for large data prediction is no longer a problem.

3. Dynamic of large data prediction. The dynamic change of the calculating factors at different point of point can cause the whole system to change and even produce the butterfly effect. If a variable is decisive for the outcome and difficult to capture, the prediction is difficult, such as human factors. Most of the scenarios for large data predictions are highly volatile but have fixed rules such as weather, stock markets, and disease. This requires the predictive system to accurately capture each variable's data and to adjust the predictions in real time. Increased data computing power outside the developed sensor network makes the above two points easier.

4. The regularity of large data prediction. The difference between large data prediction and traditional sampling based prediction is that it discovers the law between data and result based on massive historical data and real time Dynamic data, and assumes that this rule will continue, and then predict it after capturing variables. A field in itself has a relatively stable law, and large data predictions have an opportunity to be applied. The ancient night view of the sky shows that the weather is by law, so the meteorological forecast was first applied. The negative case is the unpredictable, difficult earthquake prediction of data sources, and the Shuangse lottery.

Typical application areas for large data projections

The internet provides a convenient condition for the popularization of large data prediction applications. Beyond the weather forecast, what other areas are or may be changed by large data forecasts? The following 11 areas are the most promising applications for large data prediction in combination with domestic and foreign cases.

1, sports events forecast

During the World Cup, companies such as Google, Baidu, Microsoft and Goldman Sachs have launched the results forecast platform. Baidu forecast the most bright eye, forecast the entire 64 games, the accuracy rate is 67%, enters the elimination race the accuracy rate is 94%. Now that internet companies are replacing Octopus Paul's test-water race forecasts also means that future sporting events will be dominated by big data forecasts.

Google's World Cup forecast is based on Opta QSL's massive tournament data to build its final forecast model. Baidu is the search for the past 5 years in the World 987 teams (including the national team and the team) of the 37,000 game data, while with the China Lottery web, the European index data provider Spdex data cooperation, import betting market Forecast data, A predictive model with 199,972 players and 112 million data was established, and the results were predicted.

From the experience of internet companies, as long as the historical data of sports events, and with the index companies to cooperate, you can carry out other events, such as Champions League, NBA and other events.

2. Stock Market Forecast

Last year, a study by the Warwick Business School and the Department of Physics at Boston University found that the financial keywords that users search through Google may be able to move in financial markets, with a corresponding investment strategy earning as much as 326%. Previously, some experts tried to predict stock market volatility through Twitter blog sentiment.

In theory, stock market forecasts are more suitable for the US. China's stock market can not achieve two-way profit, only the stock rose to profit, which will attract some hot money to use information asymmetry and other circumstances artificially change the rules of the stock market, so the Chinese stock market is not relatively stable law is difficult to predict, and some of the results have a decisive impact on the variable data can not be

3. Market Price forecast

The CPI represents the price fluctuations that have taken place, but the statistics are not authoritative. But big data may help people understand the future of price trends and anticipate inflation or economic crises ahead of time. The most typical case is that Ma Yun through Ali business-to-business data in advance to know the Asian financial crisis, of course, this is Ali data team credit.

The price of a single commodity is easier to predict, especially the standardized products such as air tickets, where to provide the "ticket calendar" is the price forecast, tell you a few months after the approximate price of the ticket. The production of goods, channel costs and probably gross margin in the fully competitive market is relatively stable, and price-related variables are relatively fixed, the supply and demand of goods in E-commerce platform can be real-time monitoring, so the price can be predicted, based on the results of the forecast can provide purchase time recommendations, or guide the business to carry out dynamic price adjustment and marketing activities to maximize benefits.

4, User behavior prediction

Based on user search behavior, browsing behavior, Comment history and personal data, internet business can gain insight into the overall needs of consumers, and then carry out targeted product production, improvement and marketing. "Card House" select actors and plot, Baidu based on user preferences for precision advertising marketing, Ali according to the cat user characteristics under the product line customization products, Amazon forecast user click Behavior in advance delivery is to benefit from Internet user behavior prediction.

Benefiting from the development of sensor technology and the Internet of Things, online user behavior insights are brewing. Free commercial WiFi, ibeacon technology, camera image monitoring, indoor positioning technology, NFC sensor network, queuing system, can detect the user line of mobile, stay, travel laws and other data, for precision marketing or product customization.

5, human health prediction

Chinese medicine can find some hidden chronic diseases in the body by Hango means, and even look at the constitution to know what symptoms a person may have in the future. There are some regularity in the changes of body signs, and there are some persistent abnormalities in the body before the occurrence of chronic diseases. In theory, if large data is available for such anomalies, chronic disease predictions can be made.

Combined with smart hardware, large data forecasts for chronic diseases become possible. Wearable equipment and intelligent health equipment help the network collect human health data, heart rate, weight, blood lipids, blood sugar, exercise, sleep and other conditions. If the data is accurate and comprehensive, and there are predictive models of chronic diseases that can form an algorithm, your device may in the future remind your body of the risk of a chronic disease. My Spiroo on Kickstarter can collect exhaled data from asthmatic patients to guide doctors in diagnosing their future trends. Acute disease is difficult to predict, and mutations and random features make it difficult to predict.

6. Forecast of disease epidemic situation

Based on people's search and shopping behavior to predict the possibility of large-scale outbreaks, the most classic "flu prediction" belongs to this category. If there is a growing demand for "flu" and "radix Isatidis" from a particular area, it is natural to speculate that there are flu trends.

After the World Cup, college entrance examination, attractions and City forecasts, Baidu recently launched a disease prediction products. At present, the four diseases, such as influenza, hepatitis, tuberculosis and STD, can be monitored comprehensively in every province in the country and in most of the cities and counties. In the future, the diseases predicted by the Baidu disease forecast will expand from 4 to more than 30 species, covering more common diseases and epidemics. The user can according to the local forecast result to carry on the targeted prevention.

7. Disaster forecast

Weather prediction is the most typical disaster forecast. Earthquakes, floods, high temperatures, heavy rains, such as natural disasters, if the ability to use large data can be more advanced prediction and notification will help disaster mitigation and disaster relief. Different from the past, there are dead ends in data collection, high cost problems, the era of Internet of Things can use low-cost sensor cameras and wireless communications network, real-time data monitoring collection, and then use large data prediction analysis, to achieve more accurate natural disaster prediction.

8. Environmental Change Forecast

In addition to short time microscopic weather and disaster prediction, more long-term and macroscopic environmental and ecological changes can be predicted. Forest and farmland area shrinking, wildlife plants endangered, the coastline rise, the greenhouse effect these problems are facing the Earth's "chronic problems." If humans knew more about the Earth's ecosystems and weather patterns, the more likely it would be to model future changes in the environment and prevent bad transitions from happening. Large numbers help humans collect, store, and dig more earth data, while providing predictive tools.

9. Traffic Behavior Forecast

Based on the location-based data of users and vehicles, this paper analyzes the individual and group characteristics of human vehicle travel, and carries out the prediction of traffic behavior. The traffic department can predict the flow of vehicles at different points of the road to carry out intelligent vehicle scheduling, or the use of tidal lanes; users can choose a less congested path based on the predicted results.

Baidu based on the map application of lbs forecast covers a wider range. During the Spring Festival, the trend of people's migration is to guide the setting of train lines and routes, the holiday forecast the people's choice of the scenic spots, and also the usual Baidu thermodynamic diagram tells users the city business circle, the zoo and other places of the people, guide the user travel choice and the location of the merchant.

10, energy consumption forecast

The California Grid Operations Center manages more than 80% of California's power grids, delivering 289 million megawatts of electricity to 35 million users a year, more than 25000 miles long. The center uses the space-time insight software for intelligent management, comprehensive analysis from the weather, sensors, metering equipment, such as a variety of data sources of data, forecasting the changes in energy demand around the world, intelligent power dispatching, balance the whole network of power supply and demand, and the potential crisis to make rapid response. China's smart grid is already trying to predict applications like big data.

For a single family, you can use smart home equipment, record the living habits of family members, perceive the user's comfort, predict the user's temperature-control energy demand, intelligent temperature controlling devices, and can combine the ladder Price table to help users save money. Nest a successful product based on large data to predict user energy requirements.

In addition to the more than 10 areas listed above, large data forecasts can also be applied to real estate forecasts, employment forecasts, college entrance examination scores, election results forecasts, Academy Awards forecasts, Insurance risk assessment, financial borrowers ' repayment ability assessment, etc., so that humans have a quantifiable and persuasive ability to insight into the future , the glamour of big Data forecasts is being released.