Taxi nowNow by taxi
.try it out!"'Works like a charme.
Taxi now ..... It' s better than about...faster and cheaper...and the riders rock...and the riders rock..."'The best in the city!
Now taxi driving - powered by top performance
Company accounts are a cost-effective and comfortable way to book your commercial transport needs. To call a taxi, click on the above pushbutton. As well as our immediate on-line service and our portable application, you can reach us at 301-276-5504 and our 24/7 dispatcher can receive you immediately.
TensorFlow: Predicting Taxi Needs with Tokyo Today
In February 2018, NTT DOCOMO, Japan's biggest wireless operator, started a traffic forecast services for taxi drivers. It gathers the real-time population densities of cell phone users and performs TensorFlow based analyses with a deeply adaptive modeling tool to forecast how many potential passengers could wait in each blocks or streets with an 85 to 93 percent precision in the next 30min.
Today, over 2,500 cabs in Tokyo and other large Japan metropolitan areas use the services to 1) cut each driver's mean waiting times, 2) react quickly to abrupt changes in requirements, and 3) narrow the distance between skilled and beginner riders. Advantages like these bring a significant increase in turnover for taxi companies.
Use the following information as inputs for the requirements forecasting. NewT DOCOMO gathers the site information of 60 million mobile phone users in near-redundant hours from its mobile phone networks. As NTT DOCOMO states, its taxi traffic forecasting system significantly improves precision by using its real-time demographics densities information. Following pre-processing, this information is transformed into an entry point of approximately 120 dimension.
The NTT DOCOMO researchers analyse the entrance vocabulary with two mechanical training models: Throughout the research they found that the depth study paradigm offered the best results in most areas of typically high need. However, for certain areas the statistic timeseries analyses (VAR) are better than the depth modelling.
Thus, the combined performance of the two predictive systems achieves the best overall predictive precision, covering all areas with different requirements. This forecast contains the approximate number of possible passengers in the next 30 min for each 500m x 500m cube. In the Tokyo area there are around 3,000 of them. The numbers are displayed as hot spot marks on the maps in each taxi.
Let's take a look at the depth modeling the ministry uses. A 120-dimensional entry field contains the historic recording of demographic figures, taxi activity and meteorological information, together with compiled figures such as the number of passengers on the same trip or weekday. A lot of researchers use low level RNN (Recurrent Neural Networks ) or Long Term Storage (LSTM) modeling for analyzing this kind of use.
Through the combination of the forecast from the depth modell with one from the statistic modell, the ministry achieves an error of 93-95%. The next time you come to Tokyo and never have to queue for a taxi, you might have a TensorFlow that runs on the Cloud ML engine! And, the larger the set of historic exercise information becomes, the more precise this predictive inquiry becomes basing on the actual input.
Notice: If you want to know how NTT DOCOMO gathers real-time anonymous information about demographic densities, read this document.