Research On Data Collection And Data Mining Of Wheat Growth Environment Based On The Internet Of Things

Wheat is one of three most important food crops in China. In its production process,there are effects of various environmental factors, such as light, temperature, soil,water, fertilizer, plant diseases and insect pests on the growth of wheat. How toaccomplish the real-time monitoring on environmental factors of wheat growth, andhow to realize outlier detection of wheat growth process has become an importantproblem to ensure the safe and stable production of wheat. The development of theinternet of things brings the chance for changing the traditional collection mode ofcrop growth environmental data. The wireless sensor network technology in theinternet of things can be used to achieve collection, real-time transmission anddynamic display of the wheat growth environmental data. A large number ofenvironmental data related crop growth has been accumulated with theimplementation of the internet of things based collection application of the wheatgrowth environmental data. These data record the environmental information of wheatgrowth really. But, these data has brought great difficulties for the user to find outlierinformation. Data mining is the process to find the unknown and potentially usefulinformation and knowledge from a large and noisy data, and is an effective techniqueto resolve the problem. Therefore, according to the application requirements ofreal-time data collection and outlier detection in wheat growth environmentalmonitoring, this dissertation studies two important key topics of wheat growthenvironment monitoring, i.e., data collection and outlier mining based on the internetof things and data mining technology.The major work is as follows:According to the high cost and poor efficiency problems in wheat growthenvironment data collection, this dissertation realizes the internet of things basedwheat growth environment data collection system by using ZigBee technology. Thesystem adopts ZigBee standard based wireless sensor network for a large range of wheat growth environmental data collection, such as air temperature, air humidity,soil temperature, soil humidity, wind speed, wind direction, rainfall, light radiationand so on. The system uses the ARM and Linux based embedded system deployed inthe main node on the network to package the data and do AD conversion, Focus onthe implementation of Resuming broken transfer function, and send the data to theWeb server using the3G router. The system uses the information management systemon the Web server to receive the data, and realize the functions such as data query anddownload. The design and implementation of the system is based on the actualsituation of collection of wheat growth environmental data, and then the applicationeffect of system is good.According to the application background and the complex constraints of outliermining of wheat growth environmental data, this dissertation presents a hybridclassification algorithm based on rough set and decision tree ensemble. The hybridalgorithm uses rough set as a preprocessor to effectively reduce the redundantattributes without loss of useful information, and then uses the bootstrap resamplingtechnique to generate a set of decision tree for constructing the ensemble classifier forimproving the precision of outlier mining of wheat growth environmental data. Inorder to verify the effectiveness of the algorithm, different classification methods areevaluated on wheat growth environmental data. The experimental results show thatthe proposed hybrid algorithm has improved performance.According to the problems that traditional clustering technology can not satisfy theoutlier mining application demand of wheat growth environment data, thisdissertation presents a cluster ensemble algorithm based on COD and STORM, anduses the algorithm in the outlier detection for massive and dynamic updating datastream in wheat growth environment for implementing the early warning in the wheatgrowth process control. The prototype system of outlier detection of wheat growthenvironment is designed and developed based proposed cluster ensemble algorithm.Finally, the effectiveness of the proposed cluster ensemble algorithm is verified inexperiment.The major innovations are as follows:The internet of things based data collection system of wheat growth environment isdesigned. The application effect of the system is good.A hybrid classification algorithm based on rough set and decision tree ensemble ispresented and experimental results indicate that the proposed algorithm has improved performance.A cluster ensemble algorithm based on COD and STORM is proposed. A prototypesystem of outlier detection of wheat growth environment is designed and developed.The practical application effect of the proposed cluster ensemble algorithm is verifiedin experiment.