Industrial-time-series-analysis 1.0.1

Industrial Timing Data Analysis Algorithm

Introduction

This project aims to provide a method to replicate our experiment results. We are expected to utilize Elastic Weight Consolidation (EWC) algorithm to improve the performance of multivariate time-series prediction.

Model Choice

This project provides eight algorithm to choose, including MLP / CNN/ GDN / RNN/ GRU/ LSTM/ LSTMVAE / Transformer.

Dataset Structure

• steel
  • TEst
    • Task1(拉速1.2)
      • 板柸1_9后半段数据
        • list.txt (names of multiple sensors)
        • test.csv (multivariate time series data)
    • Task2(拉速1)
      • 板柸2_4后半段数据
        • list.txt
        • test.csv
    • Task3(拉速1.4)
      • 板柸3_3后半段数据
        • list.txt
        • test.csv
  • TRain
    • Task1(拉速1.2)
      • 板柸1_1数据
        • list.txt
        • train.csv
      • 板柸1_2数据
        • list.txt
        • train.csv
      • 板柸1_9数据
        • list.txt
        • train.csv
      • 板柸1_10数据
        • list.txt
        • train.csv
      • 板柸1_11数据
        • list.txt
        • train.csv
    • Task2(拉速1.2)
      • 板柸2_1数据
        • list.txt
        • train.csv
      • 板柸2_4前半段数据
        • list.txt
        • train.csv
    • Task3(拉速1.4)
      • 板柸3_1数据
        • list.txt
        • train.csv
      • 板柸3_3前半段数据
        • list.txt
        • train.csv

Experimental Results

When you successfully complete the experiment, you should see the following figures in each task stage:

• Loss Curve for validation dataset

• Loss Curve for training dataset

• Comparison with prediction and observation

• Scatter plot with Regression Line

• Residual Distribution

Requirements

• python==3.10.2
• torch==1.12.0
• torch-geometric==2.2.0
• torch-scatter==2.0.9
• torch-sparse==0.6.14
• numpy==1.22.3
• matplotlib==3.5.2