Oil_Prices_Defense_Presentation

1.

Comparative Time Series Analysis
Brent and WTI Oil Prices
Monthly data · January 2015 — March 2026
Students: Izkhair Sanatbek, Talaspay Bulan
Forecasting project defense
01

2.

Research aim and analytical workflow
What exactly was analyzed and how the project was structured
AIM
To compare Brent and WTI oil price
dynamics and evaluate forecasting
models for monthly oil prices.
WORKFLOW
1
Data preparation
2
Exploratory analysis
3
Trend & seasonality
4
Stationarity checks
5
Forecasting models
6
Model comparison
KEY QUESTIONS
• Do Brent and WTI move together?
• Are there visible shocks or outliers?
• Is seasonality strong enough to matter?
• Which forecasting model performs better?
02

3.

Data and team contribution
Monthly observations from reliable public sources
DATASET
ROLE DISTRIBUTION
Period
Jan 2015 — Mar 2026
Frequency
Monthly
Variables
Date, Brent Spot Price, WTI Spot Price
Unit
USD per barrel
Source
U.S. Energy Information Administration
Izkhair Sanatbek
Data cleaning, descriptive statistics,
visualizations
Talaspay Bulan
Time-series models, accuracy
comparison, interpretation
Both members
Conclusion, defense preparation,
source checking
Why this data matters: Brent and WTI are global benchmarks, so their movement reflects large
market shocks and energy-price conditions.
Source note: EIA monthly spot-price data were used as the project base.
03

4.

Descriptive statistics and relationship
Initial summary shows high similarity between the two benchmarks
0.992
$66.75
$62.60
$122.71
Correlation between Brent and WTI
Mean Brent price
Mean WTI price
Maximum Brent price
SUMMARY TABLE
Indicator
Brent
WTI
Mean
66.75
62.60
Median
66.02
61.72
Min
18.38
16.55
Max
122.71
114.84
Std. Dev.
18.67
17.43
CV (%)
27.97
27.84
• Brent is consistently slightly higher than WTI.
• Both series show similar volatility: CV is around
28%.
• The very high correlation means the two
markets react to common shocks.
04

5.

Price dynamics: Brent and WTI move almost together
The strongest shocks appear around 2020 and 2022
• Both prices declined sharply during
the COVID-19 shock in 2020.
• The strongest increase occurred
around 2022.
• Brent usually remains above WTI,
but the direction of movement is
almost identical.
Main interpretation
The series are dominated by
global market shocks rather than
by a stable seasonal pattern.
05

6.

Moving average: smoothing reveals the trend
Moving average reduces short-term noise and highlights long-term movements
• Smoothing makes the 2015–2016
low-price period clearer.
• The 2020 decline is followed by
rapid recovery and growth.
• After the 2022 peak, the trend
gradually weakens.
Why it matters
Forecasting should account for
non-stable trend behavior; a
simple constant-mean model
would be weak.
06

7.

Decomposition: trend dominates seasonality
Additive and multiplicative decomposition show similar internal structure
Additive decomposition
Multiplicative decomposition
Conclusion: seasonal component exists, but major shocks and trend changes are more
important.
07

8.

Outliers and stationarity checks
Extreme values and autocorrelation explain why forecasting is not trivial
ADF TEST RESULTS
Series
p-value
Decision
Brent
0.1973
Non-stationary
WTI
0.2263
Non-stationary
• Boxplots show high-price outliers, mostly from 2022.
• ACF decays slowly, which signals strong dependence over
time.
• ADF p-values are above 0.05, so the original series are
non-stationary.
08

9.

Forecasting models applied
Four model families were compared on the same train/test split
SES
Level-only baseline
Useful when there is no clear trend.
Holt
Level + trend
Captures directional movement.
Damped
Holt
Trend with damping
Prevents unlimited trend extrapolation.
ARIMA
Autoregressive integrated model
Handles dependence and non-stationarity.
Model quality was evaluated with RMSE, MAE and MAPE. Lower values indicate better forecasting
accuracy.
09

10.

Model evaluation and residual diagnostics
Best model differs slightly by benchmark, so the conclusion must be careful
TEST ACCURACY SUMMARY
Series
Best / status
RMSE
MAPE
Brent
ARIMA
9.19
10.00%
WTI
Damped Holt
8.32
9.68%
WTI ARIMA
Close alternative
8.90
10.48%
• For Brent, ARIMA gives the lowest test error.
• For WTI, Damped Holt is slightly better by error
metrics.
• Residual check for ARIMA Brent: Ljung–Box p-value =
0.3567, so residual autocorrelation is not statistically
significant.
10

11.

Final forecast: high uncertainty remains
Forecast intervals widen because oil prices are volatile and shock-sensitive
Brent forecast
WTI forecast
Main message: the central forecast stays relatively high, but the prediction interval is wide; therefore the
forecast should be interpreted as a range, not as an exact price.
11

12.

Final conclusions
What should be remembered after the analysis
1
Brent and WTI have an extremely strong positive relationship.
2
Trend changes and market shocks are more important than seasonality.
3
Model choice must be based on test accuracy and residual diagnostics.
Limitations:
• monthly frequency only
• external shocks are unpredictable
• forecasts contain uncertainty
Defense closing line
The project shows that Brent and WTI behave as closely connected non-stationary time series;
forecasting is possible, but uncertainty must be treated seriously.
Data sources: U.S. Energy Information Administration — Europe Brent Spot Price FOB; WTI Spot Price monthly data.
https://www.eia.gov/
12