Short-term PM2.5 Prediction using Modified Attention Seq2Seq BiLSTM

In semiconductor industry, concentration of particulate matter in cleanroom is important to maintain as it can damage the wafer die in the manufacturing process. The environment in semiconductor industry is well maintained where the temperature and humidity are always stable. Hence, it is not possib...

Full description

Saved in:
Bibliographic Details
Main Authors: Krishna Yoga Utama, Ida Bagus, Tran, Duc Hoang, Jang, Yeong Min
Format: Conference Proceeding
Language:English
Subjects:
Attention Seq2Seq BiLSTM
Decoding
Electronics industry
Humidity
IoT platform
Manufacturing processes
particulate matter
PM2.5
prediction
Predictive models
Semiconductor device modeling
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In semiconductor industry, concentration of particulate matter in cleanroom is important to maintain as it can damage the wafer die in the manufacturing process. The environment in semiconductor industry is well maintained where the temperature and humidity are always stable. Hence, it is not possible to use other features to predict the PM2.5 concentrations. In this paper, we present a modified attention Seq2Seq BiLSTM model for predicting the PM2.5 concentration. Based on the proposed model, short term (60 minute ahead, 90 minute ahead, and 120 minute ahead) PM2.5 concentration predictions was built. The proposed model also compared with Seq2Seq LSTM, Seq2Seq BiLSTM, and attention Seq2Seq LSTM models where the proposed model outperformed those models by achieving lowest RMSE, MAE, and MAPE values in all prediction time length.
ISSN:2165-8536
DOI:10.1109/ICUFN55119.2022.9829659