Singapore’s limited land space is about to get worse; daily journeys by private vehicles and public transport are expected to jump 60 per cent to 15 million by 2020 from the current nine million, according to Singapore’s Land Transport Authority (LTA).
To manage travel demand, strategic land transport planning requires collecting and analysing massive information; more than 12 million records of commuter data are collected daily by the government agency.
Systems typically are not designed to crunch more than three months of data, due to limitations in processing power. At the same time, the data is insufficient for meaningful trend analysis or long-term land transport policy planning. This required a system that cross-referenced trillions of records to derive an effective land transport planning model.
With these challenges in mind, LTA embarked on one of the largest government data warehouse projects
Billions of Records
The project came to be known as Planning for Land Transport Network (PLANET). To date, PLANET has captured more than 12 million commuter records per day, which includes riders on bus and rail transit.
It supports business queries into 3.7 billion records based on two years of activity, to develop advanced predictive algorithms of travelling patterns whether by car, bus or train and the impact of traffic analytics
One key success factor for the deployment of PLANET was the prior ground work conducted by putting in place a data management programme to classify data collections and identify ownership. To establish an enterprise meta-data taxonomy and define the information structure and classification common to all within the organisation in enabling end-users’ ability to access key information easily, the Data Management Framework provided a platform for open discussion and close co-operation among user groups, resolving the conflict in usage differences in a timely manner.
“Once we identified the data and its owners, we made sure that the information only comes from the designated department. Other departments can only reference it,” said Rosina Howe-Teo, the group director and chief innovation officer, LTA.
She said many data warehousing projects fail because the same set of data was derived from multiple sources, making the information inaccurate as “different people have different interpretations of the data field.”
One Generation Ahead
Based on the performance statistics collected, PLANET achieved between 67 per cent and 99 per cent improvement over previous methods in the processing of daily data loading and transformation as well as business data queries respectively. The data load and transformation process was reduced from 12 to four hours, while business queries took 15 minutes compared to 18 hours previously.
Within the first three months of PLANET’s implementation, new capabilities have emerged. For example, 70 new analytical reports were generated for policy review, post analysis of schemes introduced and trend patterns emerged to optimise resource planning.
At a granular level, LTA can now find out about trends on bus loading, headways and even which bus stops require shelter extensions due to popular use.
Sign up for CIO Asia eNewsletters.