David Rothschild of Microsoft Research recently predicted the ultimate ownership of the Oscars by analyzing data from Oscar finalists. Who is the Oscar statuette going to spend in the end? The 85th Annual Academy Awards Gala will be held February 24 local time in the United States, the results will be announced.
You can see his latest predictions on David Rothschild's blog. As part of the prediction effort, David Rothschild worked with the Microsoft team to develop an Excel application--oscars ballot Predictor (Oscar Predictor), which provides real-time predictions for all 24 categories of Oscar winning awards.
Some of the winning probability information provided by David Rothschild:
Best Film Award: "Escape from Tehran" (Argo), the possibility of winning 93.9%.
Best Director Award: Spielberg (Steven Spielberg) (the film "Lincoln"), the possibility of winning 83.7%.
Best Actor Award: Daniel-Lewis (Daniel Day-lewis) (The film "Lincoln"), the possibility of winning 99%.
Best Actress Award: Jennifer (Jennifer Lawrence) (the "Happiness Line behind the Clouds") (Silver linings Playbook), winning 70.7%.
Of course, these figures are only preliminary forecasts, David Rothschild said: "I am sure of the predictions for some of this year's awards, but the forecast is not 100%, expecting the final results at the Oscars." ”
The probability of "Lincoln" being the best director and the best male is 83.7% and 99%, respectively.
Jennifer the probability of winning a movie Queen by the "happiness Line behind the clouds" is 70.7%
Last year, David Rothschild used a general-purpose data-driven model to correctly predict the results of the presidential elections in 50 of the 51 boroughs of 50 U.S. states and Columbia, Dist. Of, with a predicted accuracy of more than 98%.
David Rothschild said: "I predict that the Oscar-winning approach is exactly the same as predicting other things, including politics." Focus first on the most effective data and then create a statistical model that is not interfered by any particular year's results, and all models are tested and calibrated based on historical data, and we are patient when modeling to ensure that the model accurately predicts the results of the external samples, not just the results of the past. The models we create are used to predict the future, not to predict the past. "Science is the same, but proving which data is most useful is very different," he says. ”
You might think that one can conquer the votes. The forecast model of the presidential election, which is nearly 127 million unpredictable, can also be easily won in the unlikely event of a less complicated prediction, such as an Oscar vote with less than 6000 votes, but the U.S. presidential election is in stark contrast to the Oscar vote.
"I usually focus on four different types of data: polling data, forecasting market data, basic data and user-generated data," David Rothschild said. In predicting politics, I used basic data such as past election results, obligations and economic vane. Establish a baseline from the underlying data and then move on to forecast market and polling data, as these two types of data absorb and contain more electoral information. When the 2012 presidential election was predicted, I used a small number of user-generated data, but the Xbox Live data played a key role in the real-time analysis of the additional significant events. ”
"There is no need to vote for the Oscar-winning family, and the basic box office return and film rating data are not counted," he said. I'm more concerned with predicting market data, which is a major factor, and using data from some users, which helps to understand the relevance of the film's internal and different categories, such as how many awards will be won in the movie "Lincoln"? David Rothschild said.
David Rothschild stressed: "Whenever I focus on a new area, I think about key things and make sure my predictions are more meaningful." "First of all, I will determine the most pertinent forecast. As for the Oscar winners, I'm focused on winning probabilities for all 24 categories and predicting the total awards for all categories of mainstream movies. Second, all of my forecast results are updated in real time. From a research point of view, it is important to update the predictions in real time, and we can learn the value of the different events that occurred between the initial forecast and the final event. These events are the prelude to the Oscars ' attribution. Finally, I modeled on the historical data in a particular field, and then ensured the accuracy of the prediction by continually upgrading the model. I would also like to emphasize that everything we do is to ensure, as far as possible, the independence of the field and ensure the measurable nature of all problems. It would be valuable to academia and the world if the research were to produce more efficient forecasting methods and apply a wide range of problems in many fields. The