ABSTRACT
Objective To analyze the factors of errors in the pulse recognition; To improve the speed of processing massive data; To explore the method of reducing the subjective errors in pulse recognition. Methods BP algorithm based on distributed MapReduce in Hadoop environment was optimized. Optimized BP algorithm was used to self-learn pulse-sequence data to reduce fitting errors. The pulse-counting data collected by TCM electronic pulse diagnosis instrument were used as input layer of neural network. Momentum-learning rate adaptive fast BP algorithm was adopted to train neural network. Results In the training set (75%) of 768 M, a total of 35 890 data were collected, and 29 150 items were correctly predicted in stand-alone mode, with the correct rate of 81.22%. MapRedece parallel improved BP algorithm model correctly predicted 35 841 items, with the correct rate of 99.86%. Conclusion Compared with traditional BP algorithm, BP algorithm based on distributed MapReduce in Hadoop environment has smaller fitting errors, with higher accuracy.
ABSTRACT
Abstract Graph shellability is an NP problem whose classification either in P or in NP-complete remains unknown. In order to understand the computational behavior of graph shellability on bipartite graphs, as a particular case, it could be useful to develop an efficient way to generate and analyze results over sets of shellable and non-shellable instances. In this way, a sequentially designed exponential time experiment for deciding shellability on randomly generated instances was proposed in literature. In this work, with the aim of improving the performance of that experiment, we propose three alternative approaches using Apache Spark™, we called multi-core, multi-node and full-parallel. We tested and compared their execution time for bipartite graphs with 10,12,15,20 and 50 vertices with regard to the original version, and we got speedups between 1.37 and 1.67 for the first one, between 2.34 and 3.56 for the second one, and between 2.37 y 3.12 for the last version. The results suggest that parallelization could relieve the large execution times of the original approach.
Resumen La escalonabilidad* de grafos es un problema en NP del que se desconoce su inclusión en las clases de complejidad P o NP-completa. Con el fin de comprender su comportamiento computacional en el caso particular de los grafos bipartitos, podría ser de utilidad disponer de un método eficiente para generar y analizar instancias escalonables. La literatura reporta un experimento secuencial, y de costo exponencial, diseñado para determinar la escalonabilidad de un conjunto de instancias. En el presente trabajo, y con el fin de mejorar el desempeño experimento mencionado, proponemos tres alternativas utilizando Apache Spark: una multinúcleo, otra multinodo y otra completamente paralela. Además, comparamos el tiempo de ejecución de cada una de ellas respecto a la versión original en grafos bipartitos aleatorios con 10,12,15,20 y 50 vértices, y obtuvimos aceleraciones (speedups) entre 1.37 y 1.67 para la versión multinúcleo, entre 2.34 y 3.56 para la versión multinodo, y entre 2.37 y 3.12 para la versión completamente paralela. Los resultados sugieren que la paralelización del experimento podría mitigar los enormes tiempos de ejecución del enfoque original.
Resumo A shellabilidade dos grafos é um problema em NP, do qual é desconhecida sua inclusão nas classes da complexidade P ou NP-completo. A fim de compreender seu comportamento computacional no caso particular dos grafos bipartidos, poderia ser útil ter um método eficiente para gerar e analisar instâncias shellables. A literatura relata um experimento sequencial, e custo exponencial, projetado para determinar a escalabilidade do um conjunto de instâncias. Neste trabalho, e a fim do melhorar o desempenho do experimento mencionado, propomos três alternativas usando Apache Spark: uma multinúcleo, outra multinó e outra completamente paralela. Além disso, nós compararmos o tempo de execução de cada um deles respeito da versão original em grafos bipartidos com 10,12,15,20 e 50 vértices e obtivemos acelerações (speedups) entre 1.37 e 1.67 para a versão Multinúcleo, entre 2.34 e 3.56 para a versão Multinó, e entre 2.37 e 3.12 para a versão completamente paralela. Os resultados sugerem que a paralelização do experimento poderia atenuar os enormes tempos de execução da abordagem original.
ABSTRACT
Objective To establish an accurate big data platform to solve the problems due to the fusion of clinical data and genomics data.Methods The contents and technology roadmap were analyzed for establishing the accurate big data platform.The platform had its technology architecture designed with distributed memory,Hadoop technology,image fusion analysis,gene mutation point comparison and etc,and the difficulties in key technologies were discussed.Results A solution of accurate big data platform was put forward for information acquisition,storage,processing and analysis as well as infrastructure,application and etc.Conclusion The platform is of practical values for fulfilling big data network of precision medicine and research on clinical diagnosis and treatment.
ABSTRACT
Objective To establish an accurate big data platform to solve the problems due to the fusion of clinical data and genomics data.Methods The contents and technology roadmap were analyzed for establishing the accurate big data platform.The platform had its technology architecture designed with distributed memory,Hadoop technology,image fusion analysis,gene mutation point comparison and etc,and the difficulties in key technologies were discussed.Results A solution of accurate big data platform was put forward for information acquisition,storage,processing and analysis as well as infrastructure,application and etc.Conclusion The platform is of practical values for fulfilling big data network of precision medicine and research on clinical diagnosis and treatment.
ABSTRACT
Medical big data, an important strategic source of basic data for a country, will be applied in precision clinical diagnosis and treatment, decision-making support, disease monitoring, early warning and management, and public health service. The application of medical big data technology in our country is to be improved at pres-ent. How to realize the smooth transition of traditional medical data to a big data system and the added value of data by data mining and analyzing is an important problem needing to be solved immediately. The key functions, inclu-ding the general frame work and data center frame work of medical big data application information system, were thus planned and designed in this paper by constructing the regional application technology and engineering labora-tory for medical big data.
ABSTRACT
The paper introduces the research idea, design and realization of the distributed Naive Bayesian intelligent diagnosis sys-tem based on Hadoop, makes optimization and improvement according to its application in Traditional Chinese Medicine ( TCM) Hospital of Guangdong Province, including algorithm design improvement and enhancement of accuracy, extensibility and security of the system.