Time Series Analysis and Prediction of Future Commodities Prices

with SARIMA

Siyou Yao

Finance Management School, Shanghai University of International Business and Economics, Shanghai, China

Keywords: SARIMA, Future Commodities Prices, Forecast, Transaction.

Abstract: Commodities are one of the most important investing targets in the world. Forecasting the prices of

commodities precisely helps investors and corporations make reasonable transactions. This paper is based on

the prices of Brent oil, wheat, and aluminum and predicts the future prices of commodities in the energy

industry, the manufacturing industry, and the agricultural products industry. The results show that the

Seasonal Autoregressive Integrated Moving Average (SARIMA) model has a relatively excellent ability to

predict the future movements of the commodities prices. The prediction shows that all the commodities prices

will experience a decline in the next 10 months. However, the predictions are not exactly the same as the

actual movements. The fluctuation extent of the predictions is much smaller. Therefore, the SARIMA model

can help investors establish a broad idea of the future trend of the commodities prices, but it cannot help

investors do ultrashort-term trading. If the investors only focus on the trend during a longer period and ignore

making profits with the short-term fluctuations, the SARIMA model is suitable for them. In the end, this paper

suggests investors combine the fundamental analysis with the forecasting results generated by the SARIMA

model to make trading decisions.

1 INTRODUCTION

Commodities are raw materials and resources that are

normalized, tradable, and low value-added. They also

have a large volume in nature. Commodities are often

divided into three categories (Zhou, Song, & Ren,

2022): energy commodities, metal commodities, and

agricultural products commodities. As the upstream

raw material for many products, commodities have

important impacts on the development of the world

economy. Metals are unreplaceable resources and are

widely used in the manufacturing industry, oil and gas

are nonrenewable strategic resources that are the

objects of competition between the world's major

powers and leading economies, and agricultural

products are important in the food industry and can

help stabilize society (Guo, 2023). As the indicator of

the international situation, the movement of the prices

of commodities often represents macroeconomic

changes in the world, the change in supply and

demand, and the change in the market. Last but not

least, commodities can be traded by futures and

options. Investors can use commodities to earn

https://orcid.org/0009-0006-2814-2272

revenues and hedge risks. Therefore, it is meaningful

to predict the future movement of the prices of

commodities. It can help investors, companies, and

countries to make more reasonable decisions.

In 2021, Wanjuki et al. researched the Seasonal

Autoregressive Integrated Moving Average

(SARIMA)’s ability to forecast the price index of

food and beverages in Kenya (Wanjuki, Wagala, &

Muriithi, 2021). After comparing all the models with

different parameters, it turned out that SARIMA (1, 1,

1) (0, 1, 1)12 was the best fit for the data in the study.

In 2022, Li et al. applied the ARIMA model to gold

prices and gave investing suggestions based on the

forecasting results (Li, 2022). The study encouraged

investors to combine international situations with the

results. In 2023, ARIMA and SARIMA’s ability to

predict future crude oil prices were compared by

Ariyanti, and it turned out that both models were

excellent in forecasting (Ariyanti & Yusnitasari,

2023), which meant ARIMA and SARIMA models

were suitable for time series analysis of crude oil

prices. Gasper et al. conducted a study on forecasting

crude oil prices in Tanzania using the ARIMA model

Yao, S.

Time Ser ies Analysis and Prediction of Future Commodities Prices with SARIMA.

DOI: 10.5220/0013699500004670

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 2nd International Conference on Data Science and Engineering (ICDSE 2025), pages 467-475

ISBN: 978-989-758-765-8

467

(Gasper & Mbwambo, 2023). This study found out

that even under the serious fluctuation caused by the

conflict in Ukraine and the coronavirus outbreak, the

ARIMA model could still capture the potential

movement in crude oil prices. In 2024, Guzman

affirmed ARIMA’s ability to predict future corn

prices in Mexico (Guzma, 2024). This research also

emphasized the importance of forecasting future

prices of agricultural products by pointing out that

future corn prices would influence farmers' interests

and sustainable agricultural development. Bagrecha

et al. used the ARIMA model to forecast silver prices

in India. However, the results only explained 26% of

the observed silver price changes. It suggested that

the ARIMA model was too simple for silver price

prediction and that more factors should be taken into

consideration (Bagrecha et al., 2024). In 2025, Ojha

et al. utilized the SARIMA model to predict global

wheat prices (Ojha & Karki, 2025). They pointed out

its importance in helping investors and countries

make reasonable strategic decisions. However, the

study claimed that the prediction is only suitable for

short-term analysis, and many external factors are not

considered. The results may not fit other commodities

and periods.

This paper aims to improve the ability to predict

the future prices of different commodities with the

SARIMA model, hoping to expand from the price

prediction of the commodity to the prediction of the

future development trend of the industry

corresponding to that commodity, and finally explore

the changes in industries and changes in the world

economy. This study first searches the prices of the

commodities in each of the three categories

mentioned above. Based on the different

characteristics of all the commodities in each

category, this study finally chooses aluminum in

metals, wheat in agricultural products, and Brent oil

in energy as the data to be researched. This paper first

applies pre-processing procedures to the data and then

uses the SARIMA model to fit the data and make

forecasts of future data in the next 10 months. The

differences between the future data and the actual

data are compared to evaluate the ability of the

SARIMA model. Conclusions are reached based on

the results and international events during the

research period.

2 DATASETS

2.1 Data Collection and Description

This paper extracts the datasets from Kaggle, and the

original datasets are extracted from Alpha Vantage

API using Python. The dataset contains monthly

historical prices of 10 different commodities from

January 1990 to March 2023. Prices are reported in

USD per unit of measurement for each commodity.

The prices of aluminum, wheat, and Brent crude oil

were selected in the study.

Crude oil is one of the most important energies in

the world. It can represent the energy industry. This

paper selects the price of the Brent crude oil which is

a blended crude stream produced in the North Sea

region. This is because Brent crude oil is one of the

most important crude oil pricing benchmarks in the

world. It is widely used in international oil price

quotation and contract settlement, and it serves as a

marker for pricing a number of other crude streams.

Compared with OPEC and WTI prices, the price of

Brent crude oil is more comprehensive and

transparent, because it is not affected by the US

domestic factors and the political factors of the

Organization of the Petroleum Exporting Countries

(OPEC).

Aluminum is widely used in aerospace,

automotive, construction, packaging, and many other

manufacturing and industrial sectors. It can represent

the manufacturing industry. The price movement of

aluminum directly reflects the economic performance

and industry cycles. Compared to gold and silver,

aluminum is less used as a precious metal and

investment target. Therefore, its price volatility is

usually less influenced by macroeconomic

uncertainty and safe-haven demand, which makes it

more appealing data in this study.

Wheat is one of the world's leading food crops. It

is widely used in food production and is important in

global agricultural trade. It can represent the

agricultural industry. Compared with other

agricultural products, wheat has a wider application.

For example, it can be processed into bread and beer.

2.2 Data Pre-processing

Since the characteristics of the prices of different

commodities are similar, the data pre-processing is

the same. Therefore, this paper will only focus on the

overall procedure in this part.

First of all, this paper generates the time series

plots of the prices of three commodities, the results

are shown in Figures 1, 3, and 5. From the plots, it

should be decided whether the prices of the three

commodities are stationary. This paper then applies

the log transformation to the data to make them

stationary. This paper also uses STL decomposition

to detect the characteristics of the data. The results are

shown in Figures 2, 4, and 6. From the four plots

ICDSE 2025 - The International Conference on Data Science and Engineering

468

generated by the STL decomposition, the seasonality,

trend, and volatility can be checked. Then, it can be

decided which model to use.

Figure 1: The Time Series Plot of the Brent Oil Prices after

Log Transformation. (Picture credit: Original)

Figure 2: The STL Decomposition on Brent Oil Prices.

(Picture credit: Original)

Figure 3: The Time Series Plot of the Wheat Prices after

Log Transformation. (Picture credit: Original)

Figure 4: The STL Decomposition on Wheat Prices.

(Picture credit: Original)

Figure 5: The Time Series Plot of the Aluminum Prices

after Log Transformation. (Picture credit: Original)

Figure 6: The STL Decomposition on Alunimun Prices.

(Picture credit: Original)

From the time series plot of the Brent oil prices,

wheat prices, and aluminum prices, it is obvious that

the data is not stationary even after the log

transformation. It means that the data need a

differencing process. According to the decomposition,

all the data show a slightly upward trend and

seasonality. The remainders show that the

Time Series Analysis and Prediction of Future Commodities Prices with SARIMA

469

characteristics of the data have been nearly fully

extracted by the trend and seasonality.

Therefore, this paper generates the ACF plot and

the PACF plot to further check whether the data are

stationary. The results show that the autocorrelation

of all three datasets has a downward trend as the lag

increases. In addition, the PACF values of the data are

extremely large when the lag equals one. All in all, it

means that the data are not stationary even after the

log transformation. The data may need differencing.

To further ensure whether the data need

differencing, this study also utilizes the Augmented

Dickey-Fuller unit root test as the stationary test. The

test can identify stationary data and judge whether the

data needs differencing. If the p-value is bigger than

0.05, the test will reject the null hypothesis that the

time series data is not stationary and has a unit root.

The results are shown in Table 1.

Table 1: The Results of ADF Tests on Commodities Before

Differencing.

Commodity

ADF test before differencing

Oil

p-value = 0.3696

Wheat

p-value = 0.4663

Aluminum

p-value = 0.03145

From Table 1, the p-values of the Brent oil prices,

the wheat prices, and the aluminum prices are 0.3696,

0.4663, and 0.03145 respectively. It can be seen that

the p-values of the Brent oil prices and the wheat

prices are much larger than 0.05, while the p-value of

the aluminum prices is less than 0.05. Combined with

all the results reached above, this research conducts a

first-order differencing to all the data.

After the differencing, both the ACF and PACF

plots and the ADF test are utilized again to decide

whether the data needs more processing procedures.

The results are shown in Figures 7, 8, 9, 10, 11, and

12 and Table 2.

Figure 7: The ACF Plot of the Brent Oil Prices after