Tentative Schedule: 22 November 2017
Morning Session: 9.00 A.M. -12.00 A.M.
Topic: An Overview of IoT Middleware
Dr. Tuul Triyason
Dr. Debajyoti Pal
Researchers@IP Communications Laboratory (I-Lab)
School of Information Technology, King Mongkut's University of Technology Thonburi.
The Internet of Things (IoT) has become one of the fastest growing enterprise technology trends in the last few years. From productivity wearables to extremely sophisticated industrial deployments of sensors, enterprise IoT solutions are dominating the technology agenda of modern enterprises. The emergence of enterprise IoT has brought together a new set of integration challenges connecting the new world of smart devices with existing line of business systems. The integration challenges created by enterprise IoT topologies have been unprecedented in the enterprise. Never before, have companies encountered integration scenarios involving such a large number of endpoints, large volumes of data, and heterogeneous environments. Quickly, enterprises are discovering a new reality: IoT requires a new type of middleware. This tutorial session focuses on the concept of IoT middleware and its applications. The organizer team will introduce a comprehensive overview of Netpie, the first IoT middleware platform of Thailand. NETPIE platform is a cloud-based platform-as-a-service that facilitates interconnecting IoT devices (“things”) together in a most seamless and transparent manner possible by pushing the complexity of connecting IoT devices from the hands of application developers or device manufacturers to the cloud.
- Dr. Tuul Triyason, An Overview of IoT middleware, 30 min;
- Dr. Debajyoti Pal, An Introduction to Netpie, The First IoT Platform of Thailand, 150 min;
Afternoon Session: 1.30 P.M. – 4.30 P.M.
Topic: Big Data Analytics: Large Scale Recommender Systems
Dr. Praisan Padungweang
Big Data Specialist
School of Information Technology,
King Mongkut's University of Technology Thonburi.
The recommender systems play an important role for companies that have adopted data-driven strategies. These systems can produce personalized recommendation for each customer. Frequent pattern mining and corroborative filtering are wildly used as recommendation engine. The former can be used when user history of transactions are available, e.g., purchase history, web access history. The latter can be used when product ratings are available, either implicit or explicit rating. Consequently, companies need to gather and keep as much customer data as possible. The amount of data tends to increase over time. Some of analytical processing cannot be done in acceptable time period using community hardware. Although, high computing hardware such as mainframe computer is one of option for performance scaling, it is a high-cost system, software, hardware and maintenance. Therefore, Hadoop and its ecosystem, using open source software and community hardware, become a top choice providing good performance at reasonable cost. This tutorial focuses on both theory and implementation of recommender systems using apache Spark on Hadoop ecosystem. Firstly, a scalable association rules mining will be discussed and applied to raw transactions data. Secondly, a model based corroborative filtering will be considered and applied to user's product ratings dataset. Model evaluation, selection and deployment are also discussed and implemented. This tutorial uses the latest technology based on Hadoop environment including storing data in Hadoop distributed file system (HDFS), querying using HiveQL, pre-processing using key-value style, bringing computation to data, using resilient distributed dataset (RDD) and Spark machine learning library.