With China's sustained rapid economic development, China's urban rail transit construction market in the peak in the next long period of time. According to incomplete statistics show that in 2017, Beijing, Guangzhou, Shenzhen, Nanjing, Suzhou, Qingdao, Hangzhou and other 30 new 63 urban rail transit lines will be started soon.
Increasing demand for urban rail transit, the corresponding subway project, the metro rail project also will demand surge. Since the track construction has a long duration, high investment, high operating costs and other characteristics, it is necessary to track passenger flow forecast, to analyze and evaluate all angles of the economic, environmental and social From a quantitative point of view, in order to evaluate viable rail construction projects and implementation of benefits. The dynamic changes in passenger flow of real-time tracking and analysis system, traffic control system is a prerequisite for changes of work organization rail transport is run smoothly.
Metro this "underground city" to make every moment of interconnection between people, vehicles, roads, environment, resulting in a massive complex data in the course of operations, such as personal behavior, timing, subway riders stability, ride the subway distance, site information, travel costs, these metro data will be an important basis for us to predict trends in passenger rail.
Personalized travel forecasting analysis
Here we have an example to talk about the Beijing subway typical application scenarios - predictive analytics personalized travel.
Complexity and the complexity of the Beijing subway passenger in the country topped the million passengers daily traffic has become the norm. Since the data traffic is relatively large-scale subway system, if done overall macroeconomic analysis based on traditional OD data, we can build a relatively simple prediction model, but poor accuracy; it is necessary to consider whether the existing big data analytics technology to travel on individual records predictive analysis, so that the traffic forecasting process more efficient, but also more accurate predictions.
We will be conducted through SaCa RealRec scientific data platform for sample data analysis, feature extraction, user modeling, realization of individual card users out of the station for real-time accurate prediction function, which can make the macro forecast passenger flow forecasting a site on an individual basis.
FIG above procedure, we chose a decision tree model to predict card users stop sites, the main process is as follows:
the original data feature extraction
multidimensional feature analysis based SaCa RealRec assembly platform provides scientific data, can easily filter the raw data recorded is insufficient user, and wherein successive discrete values of the processing, such as discrete "time" feature.
Construction of multi-dimensional feature vector
of individual travel behavior has a strong user personalization features of the law, based on scientific data platform SaCa RealRec powerful distributed computing capabilities, you can build predictive models personalized for each card user, and ultimately large-scale reinforcement learning model.
For building machine learning models complete, SaCa RealRec scientific data platform system for evaluating the prediction error of the decision tree analysis of the situation. The evaluation results of the feedback, the operator can guide the art to optimize the process of constructing a decision tree, pruning of the decision tree to tree node by setting a threshold of confidence, SaCa RealRec scientific platform enables user data pit prediction after pruning 97.9% accuracy rate, greatly improving the user's ability to predict travel behavior. 
In the above scenario, relying on scientific data platform SaCa RealRec travel multidimensional data analysis dig mass transit users to extract relevant factors stint with the user, and using the data modeling features for users to travel through machine learning algorithms to stop site for users to achieve real-time monitoring to predict.
Subway travel in the field of rail transportation, there are many scenes wait for the use of Big Data technologies to solve, such as helping customers more efficiently manage the operation and maintenance of rail transportation planning, efficient rail transportation of high throughput operation, increased delivery of public transport the ability to provide the public with more convenient and satisfactory travel programs.
Take full advantage of big data mining, data analysis and visualization technology to better build intelligent rail transportation, will promote the rapid development of urban intelligent transportation, and create endless possibilities to ease urban traffic congestion, improve traffic environment.
About RealRec SACA
SACA RealRec scientific data platform for big data intelligence strategy of focusing high-level analysis and forecasting services platform, based on large-scale machine learning algorithm library data and other science-related technology to improve the capacity and efficiency of enterprise build intelligent applications, simplify complex machine learning algorithms cost, the business model to help companies achieve data-driven. More scientific data to understand the content platform