According to Baidu’s travel forecast results, we’ve identified the top ten most popular scenic spots in advance. Iconic destinations like Mount Tai and Jinggangshan need no introduction. Interestingly, there are also many lesser-known spots that are actually extremely crowded. Baili Azalea, for instance, is still relatively obscure in the public eye. In reality, it boasts the reputation of being “the world’s largest natural garden.” On the eve of last year’s and this year’s Tomb-Sweeping holiday, the daily visitor count at Baili Azalea exceeded 200,000, far surpassing the scenic area’s capacity and resulting in severe congestion. Other traditional hotspots, such as Jiuzhaigou and Mount Wutai, remain just as popular during this year’s Tomb-Sweeping holiday. There’s also Sand Lake in Ningxia, named by CNN as one of “China’s 40 Most Beautiful Scenic Spots” and “China’s Top Destination for Bird Watching.” This national 5A-level scenic area is also extremely popular. During this year’s Tomb-Sweeping period, Sand Lake is hosting its 3rd International Bird-Watching Festival, which promises to be bustling with activity.
So, how does Baidu Prediction uncover these insights that used to rely on fortune-tellers’ guesses? In fact, the core of the travel forecast is a predictive model based on Baidu’s big data. On one hand, Baidu has recorded historical crowd data for all scenic spots across the country through its LBS products. On the other hand, from Baidu’s search logs, it gathers demand data for any attraction users wish to visit, as well as information about upcoming major events, folk gatherings, and other relevant activities in those areas. It also obtains corresponding weather and air quality data for those times. These data points are typically a sequence of values arranged over time, known as a “time series,” or simply “timeseries.” Baidu’s Big Data Department, in collaboration with Baidu’s Institute of Deep Learning (IDL), has developed a timeseries prediction model called the “state space model” for the travel forecast product. The so-called “states” refer to various factors that influence the timeseries. The effects of these factors on the timeseries are quantified and systematized within the model, enabling future predictions. This model can easily incorporate new factors as additional states to improve prediction accuracy.
Beyond scenic spot predictions, Baidu Big Data Forecast will also offer predictions for tourism city popularity, major disease incidence rates, college entrance exam admissions, World Cup outcomes, movie box office performance, real estate prices, and more. Prediction is the core value of big data. Baidu aims to explore, analyze, and forecast using big data to seek its application in the realm of people’s livelihoods, helping the public access information and make life decisions.