Real-Time Interpretation And Optimization Of Stream Time Series In Big Data

In the futures program trading environment,the futures stream time series is the data of price obtained from different commodity contracts.The time series data has characteristic of large,high latitude and changes constantly over time.It becomes very important to deal with the massive historical data and real-time data flow in the process of program trading.The interpreter combines the data and explains the trading model or trading strategy into the machine language by using computer and other technologies.Finally the interpreter helps traders to complete models and gains benefits.Based on this problem,this thesis proposes a real-time interpretation model of the stream time series.The model consists of the data acquisition module,the real-time processing module and the interpreter module.The main research work is as follows:The first,in the data acquisition module,the traditional database can't well storage the massive historical data and the real-time stream.Therefore,this thesis combines with the advantage of the Hadoop distributed storage system and puts forward a storage method based on Hadoop cloud platform.The HBase database storages the data,then the Flume framework collects data of each node in real-time,finally the message queue of Kafka distributes the message to the next phase.Then,in the real-time processing module,this thesis proposes the ARTMMR algorithm that combines the traditional Apriori algorithm with the batch processing MapReduce framework of the Hadoop cloud platform.The ARTMMR algorithm uses the pipeline technology to meet the real-time environment.At last,the experimental results show the good performance of the ARTMMR algorithm.Finally,it becomes the main research that the program trading system uses the C-like language as the model description language.Therefore,this thesis combines the characteristic of the program trading and develops an interpreter of the MVC pattern.The interpreter can interpret real-timely the trading strategies or models written by traders and return the corresponding performance report for optimizing the model and gaining profits.As the same time,the multi-threading technology allows traders to log in different accounts and choose a variety of contract trading transactions for increasing the trading opportunities.