Air Quality Monitoring and Analysis with Automatic Forecasting Using Machine Learning
碩士 === 東海大學 === 資訊工程學系 === 106 === The time series prediction problem is a challenging case study of predictive modeling because time series increase the time-dependent convolution between input variables. However, Recurrent Neural Network (RNN) is a powerful neural network that is not only memory b...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2018
|
Online Access: | http://ndltd.ncl.edu.tw/handle/j77pg4 |
id |
ndltd-TW-106THU00394012 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-106THU003940122019-05-16T00:22:55Z http://ndltd.ncl.edu.tw/handle/j77pg4 Air Quality Monitoring and Analysis with Automatic Forecasting Using Machine Learning 運用機器學習分析於自動化預測空氣品質監測 LEE, CHING-FANG 李靜芳 碩士 東海大學 資訊工程學系 106 The time series prediction problem is a challenging case study of predictive modeling because time series increase the time-dependent convolution between input variables. However, Recurrent Neural Network (RNN) is a powerful neural network that is not only memory but also capable of processing time series-related data. In this paper, we use RNN to analyze the air pollution PM2.5 and its automated prediction. Moreover, in the experiment, we established a distributed computing environment based on RHadoop and analyzed the air pollution. In addition to accessing timely data through the MySQL database, Sqoop is also used to access HBase historical data quickly. Also, we also discussed and experimented with a missing value of data to find a complementary method that does not affect prediction accuracy or enhance prediction accuracy. Moreover, in the experiment, we use the average absolute error percentage (MAPE) value to quantify the short-term prediction accuracy of PM2.5 and control the MAPE by 0.2 to 0.5 intervals. Moreover, without affecting the accuracy, experiment, and tune for each parameter of RNN, and further, develop RNN automated training program. Finally, because of the help of the visualization for developers and users, we also use R and Shiny to visualize the RNN training results and help optimize the parameters of the RNN training module so that developers can quickly analyze the information and The predicted PM2.5 value is displayed on the map for user reference. YANG, CHAO-TUNG 楊朝棟 2018 學位論文 ; thesis 98 zh-TW |
collection |
NDLTD |
language |
zh-TW |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 東海大學 === 資訊工程學系 === 106 === The time series prediction problem is a challenging case study of predictive modeling because time series increase the time-dependent convolution between input variables. However, Recurrent Neural Network (RNN) is a powerful neural network that is not only memory but also capable of processing time series-related data. In this paper, we use RNN to analyze the air pollution PM2.5 and its automated prediction. Moreover, in the experiment, we established a distributed computing environment based on RHadoop and analyzed the air pollution. In addition to accessing timely data through the MySQL database, Sqoop is also used to access HBase historical data quickly. Also, we also discussed and experimented with a missing value of data to find a complementary method that does not affect prediction accuracy or enhance prediction accuracy. Moreover, in the experiment, we use the average absolute error percentage (MAPE) value to quantify the short-term prediction accuracy of PM2.5 and control the MAPE by 0.2 to 0.5 intervals. Moreover, without affecting the accuracy, experiment, and tune for each parameter of RNN, and further, develop RNN automated training program. Finally, because of the help of the visualization for developers and users, we also use R and Shiny to visualize the RNN training results and help optimize the parameters of the RNN training module so that developers can quickly analyze the information and The predicted PM2.5 value is displayed on the map for user reference.
|
author2 |
YANG, CHAO-TUNG |
author_facet |
YANG, CHAO-TUNG LEE, CHING-FANG 李靜芳 |
author |
LEE, CHING-FANG 李靜芳 |
spellingShingle |
LEE, CHING-FANG 李靜芳 Air Quality Monitoring and Analysis with Automatic Forecasting Using Machine Learning |
author_sort |
LEE, CHING-FANG |
title |
Air Quality Monitoring and Analysis with Automatic Forecasting Using Machine Learning |
title_short |
Air Quality Monitoring and Analysis with Automatic Forecasting Using Machine Learning |
title_full |
Air Quality Monitoring and Analysis with Automatic Forecasting Using Machine Learning |
title_fullStr |
Air Quality Monitoring and Analysis with Automatic Forecasting Using Machine Learning |
title_full_unstemmed |
Air Quality Monitoring and Analysis with Automatic Forecasting Using Machine Learning |
title_sort |
air quality monitoring and analysis with automatic forecasting using machine learning |
publishDate |
2018 |
url |
http://ndltd.ncl.edu.tw/handle/j77pg4 |
work_keys_str_mv |
AT leechingfang airqualitymonitoringandanalysiswithautomaticforecastingusingmachinelearning AT lǐjìngfāng airqualitymonitoringandanalysiswithautomaticforecastingusingmachinelearning AT leechingfang yùnyòngjīqìxuéxífēnxīyúzìdònghuàyùcèkōngqìpǐnzhìjiāncè AT lǐjìngfāng yùnyòngjīqìxuéxífēnxīyúzìdònghuàyùcèkōngqìpǐnzhìjiāncè |
_version_ |
1719166099870711808 |