A Two-Stage Method for Commodity Price Trend Forecasting SIGIR - - PowerPoint PPT Presentation

▶

Nov 10, 2023 385 likes •503 views

A Two-Stage Method for Commodity Price Trend Forecasting SIGIR Workshop: FinIR 30 #$ July 2020 ustc_youdu: Bingjie Liang, Huixin Liu and Chujing He University of Science and Technology of China Problem Description The main task is to build

SLIDE 1

A Two-Stage Method for Commodity Price Trend Forecasting

SIGIR Workshop: FinIR 30#$ July 2020 ustc_youdu: Bingjie Liang, Huixin Liu and Chujing He University of Science and Technology of China

SLIDE 2

The main task is to build prediction models and use data from 2003 to 2018 to predict the six metals’ price movement direction in 2019 at three time-horizons.

Problem Description

SLIDE 3

Time series data
Daily transaction data in LME for six metals
Daily transaction data of main relevant commodities

and financial indices

Textual data
Analyst Reports published by institutional trader and

News Reports, collected from both English and Chinese sources.

Data Description

SLIDE 4

Data Preprocessing

11 11 11 1

SLIDE 5

3D
0E91H

Opening price

Highest price

.L91H

Lowest price

CI91H

Closing price

5CKD

Trading volume

Open interest
67

67E

/7E

228 E

46 E

56E

216 E

26 E

1HHECCHN .C9 1HHECCHN .C9 1HHECCHN

SLIDE 6

Feature Selection

SLIDE 7

Feature Selection
Use different feature combinations to train traditional

classification models

– Naïve BayesKNNRandom ForestSVM

Input of classification model

– Feature values of the current trading day and the expected trading day

Important features

– Open_PriceHigh_PriceLow_PriceClose_Price

Current day Expected day Classification model

Input Output

Trend label: 0/1

SLIDE 8

Model Structure

Close_Price(lag=5)

!"#$%&'( ⋯ !"#( !* LSTM Network

!*'+, !*'-,

Close_Price_Pred Close_Price Close_Price_Pred_1d Close_Price Close_Price_Pred_20d Close_Price Close_Price_Pred_60d

Random Forest First Stage: Prediction LGB Classifier Random Forest

1d label: 0/1 20d label: 0/1 60d label: 0/1

Second Stage: Classification

normalize

!*'(

!"#$%&'( ⋯ !"#( !* !"#$%&'( ⋯ !"#( !* !"#$%&'( ⋯ !"#( !*

Length = lag

SLIDE 9

Prediction Curve

Figure: Prediction curves for Aluminum on validation set

SLIDE 10

Results

Total Task1(1d) Task2(20d) Task3(60d) 55.18225736 50.00000000 48.74835310 66.79841897

SLIDE 11