---
title: ARIMA Models for Time-Series Forecasting
url: https://www.emergentmind.com/topics/arima
type: topic
---

# ARIMA Models for Time-Series Forecasting

ARIMA, the Auto Regressive Integrated Moving Average model, is a classical linear framework for univariate time-series forecasting in which a nonstationary series is differenced until it becomes approximately stationary and the resulting process is modeled with autoregressive and moving-average terms. In the papers surveyed here, ARIMA is consistently presented through the Box–Jenkins workflow and parameterized as ARIMA\((p,d,q)\), where \(p\) is the autoregressive order, \(d\) is the differencing order, and \(q\) is the moving-average order [1803.06386]. A standard compact form used in several of these studies is
\[
\phi(B)(1-B)^d Y_t = \theta(B)\varepsilon_t,
\]
with \(\phi(B)\) the AR polynomial, \(\theta(B)\) the MA polynomial, and \((1-B)^d\) the nonseasonal differencing operator [1902.07953].

## 1. Formal structure and notation

ARIMA is usually introduced as a generalization of ARMA that combines three components: autoregression, integration through differencing, and moving average [1803.06386]. In the stationary case, the component models are written as
\[
x_t = c + \sum_{i=1}^{p} \phi_i x_{t-i} + \varepsilon_t
\]
for AR\((p)\), and
\[
x_t = \mu + \sum_{i=0}^{q} \theta_i \varepsilon_{t-i}
\]
for MA\((q)\), with ARMA\((p,q)\) combining both terms [1803.06386]. ARIMA extends this by applying differencing before the ARMA stage.

Several papers also give equivalent differenced-domain formulations. One representative form is
\[
\Delta^d Y_t = \sum_{i=1}^{p}\phi_i \Delta^d Y_{t-i} + \sum_{j=1}^{q}\theta_j \varepsilon_{t-j} + \varepsilon_t,
\qquad \varepsilon_t \sim N(0,\sigma^2),
\]
which makes explicit that the modeled object is the differenced series rather than the original level process [2502.07841]. Another common non-operator form is
\[
x_t = \sum_{i=1}^{p} \phi_i x_{t-i} + a_t - \sum_{j=1}^{q} \theta_j a_{t-j},
\qquad a_t \sim NID(0,\sigma_a^2),
\]
used in a temperature-forecasting comparison [2011.04452].

Seasonal structure is handled by the multiplicative extension
\[
\text{ARIMA}(p,d,q)\times(P,D,Q)_S,
\]
where \(P,D,Q\) are the seasonal AR, differencing, and MA orders and \(S\) is the seasonal period [1803.06386]. In electricity-price forecasting, the seasonal model is written as
\[
\phi_p(B)\,\Phi_P(B^s)\,\nabla^d \,\nabla_s^D\, y_t
=
\mu + \theta_q(B)\,\Theta_Q(B^s)\,\varepsilon_t,
\]
with \(s=24\) for hourly daily seasonality [1801.02485].

This formalism situates ARIMA as a linear, lag-structured, stationarity-adjusted forecasting model. Several papers describe it as a standard econometric benchmark, especially for short-horizon or one-step-ahead prediction tasks in which lag interpretability and explicit differencing remain desirable [1803.06386].

## 2. Identification, stationarity, and estimation

The methodological core of ARIMA in these papers is the Box–Jenkins sequence: model identification, parameter estimation, residual diagnostic checking, and forecasting [2011.04452]. Identification begins with stationarity assessment. Time plots, ACF, PACF, the Augmented Dickey–Fuller test, and sometimes KPSS are used to determine whether differencing is required [1902.07953].

The first-difference operator is repeatedly written as
\[
\nabla Y_t = Y_t - Y_{t-1},
\]
or, in one paper’s notation,
\[
y_t = x_{t+1} - x_t,
\]
with higher-order differencing applied if the once-differenced series remains nonstationary [1902.07953]. Variance stabilization is sometimes added before differencing. The Box–Cox transformation appears in the form
\[
W_t = \frac{Y_t^\lambda - 1}{\lambda}, \qquad \lambda \neq 0,
\]
and
\[
W_t = \log Y_t, \qquad \lambda = 0,
\]
in malaria incidence forecasting, where differencing alone removed the unit root but produced a model with one non-significant coefficient and non-normal residuals, whereas Box–Cox transformation before differencing led to an acceptable ARIMA\((2,1,0)\) model [1902.07953].

Order identification is typically tied to ACF and PACF. One common rule in these papers is that PACF is used to determine \(p\) and ACF to determine \(q\), often in combination with information criteria such as AIC or BIC [1902.07953]. In the Ghana insurance-penetration study, `auto.arima(..., ic="aic", trace=TRUE)` searched a candidate set that included both seasonal and nonseasonal models and selected ARIMA\((3,1,0)\) because it had the lowest AIC, \(-398.502\) [2502.07841]. In the Philippine malaria study, three tentative models were considered after stationarity treatment,
\[
\text{ARIMA}(0,1,2),\quad \text{ARIMA}(2,1,0),\quad \text{ARIMA}(2,1,2),
\]
and ARIMA\((2,1,0)\) was chosen because it had the smallest AIC, \(296.80\) [1902.07953].

Estimation is usually carried out by maximum likelihood. This is stated explicitly in the malaria paper, the Ghana insurance study, the Noja water-pressure study, and the Italian COVID-19 hospitalization comparison [1902.07953]. A plausible implication is that, across contemporary applications, ARIMA practice is increasingly software-mediated: several studies rely on `auto.arima`, `statsmodels`, EViews, or SPSS rather than manual Yule–Walker-style workflows.

## 3. Diagnostics, residual structure, and probabilistic forecasts

The standard adequacy condition stated across multiple studies is that residuals should behave like white noise. One paper defines residuals as
\[
a_t = y_t - \hat{y}_t,
\]
and notes that, for a satisfactory model, residuals should be uncorrelated and have zero mean [2011.04452]. In the malaria study, residual adequacy was checked with residual plots, Q–Q plots, residual ACF/PACF, and the Ljung–Box statistic
\[
Q^* = n(n+2)\sum_{k=1}^{m}\frac{r_k^2}{n-k},
\]
with the final ARIMA\((2,1,0)\) model yielding \(Q^*=4.1824\) and \(p=0.8403\), supporting the white-noise interpretation [1902.07953].

The Ghana study reports a similar residual-validation pattern for ARIMA\((3,1,0)\): Ljung–Box \(Q^* = 5.1367\), \(p=0.3994\), residual ACF and PACF within standard error limits, and residual stationarity supported by ADF, PP, and KPSS tests [2502.07841]. In the Noja water-supply application, the final ARMAX\((4,3)(0,0)\) model also passed residual checking, with Durbin–Watson
\[
DW = \frac{\sum_{t=2}^{n}(e_t-e_{t-1})^2}{\sum e_t^2}
\]
reported as \(1.999668\), effectively indicating no first-order residual autocorrelation [2512.09717].

Forecast assessment in these papers uses a wide range of criteria. Mean Squared Error appears as
\[
\mathrm{MSE}=E[(Y-\hat{Y})^2]
\]
in temperature forecasting [2011.04452], while RMSE is written as
\[
\text{RMSE}=\sqrt{\frac{1}{N}\sum_{i=1}^{N}(x_i-\hat{x}_i)^2}
\]
in economics and finance [1803.06386]. MAPE, MASE, Theil’s \(U\), and Gaussian-white-noise checks on forecast errors also recur [2010.11617].

A notable theoretical extension derives the joint distribution of multistep ARIMA forecast errors rather than only marginal one-step variances. Under Gaussian assumptions, the paper “Joint Probability Distribution of Prediction Errors of ARIMA” shows that the vector
\[
\{Err(X_{n+1}),Err(X_{n+2}),\dots,Err(X_{n+h})\}
\]
is jointly multivariate normal, with covariance obtained through explicit linear transformations of the underlying white noise [1811.04685]. This shifts ARIMA from pointwise forecasting toward horizon-wide probabilistic monitoring.

## 4. Variants and extensions

The surveyed literature treats ARIMA less as a single fixed model than as a family of related constructions. Seasonal ARIMA is the most classical extension; ARIMA with exogenous regressors appears as SARIMAX, ARMAX, or ARIMAX when external drivers are added [1801.02485]. In the Noja hydraulic-control study, the final model is not plain ARIMA but
\[
\text{ARMAX}(4,3)(0,0),
\]
with contemporaneous exogenous inputs \(P2_t\) and \(FLOW_t\), and fitted coefficients \(FLOW=-0.059023\) and \(P2=1.020825\), reflecting hydraulic head-loss and control effects [2512.09717].

In electricity-price forecasting, the key object is not raw RTLMP but the differential series
\[
ALMP_t = DALMP_t - RTLMP_t.
\]
A forecast of this spread is converted back to a real-time price forecast via
\[
\widehat{RTLMP}_{t+i} = DALMP_{t+i} - \widehat{ALMP}_{t+i}.
\]
On this transformed target, ARMAX\((1,1)\)-GARCH\((1,1)\) with a weekday/weekend indicator outperformed the paper’s other ARIMA-family models [1801.02485]. This suggests that, in some domains, exogenous structure and variance modeling are more consequential than the base ARIMA order alone.

Hybridization is a more recent direction. In one algorithmic-investment study, ARIMA is not the final predictor but a feature-generating component whose residuals are passed to an LSTM together with closing price, volatility, and trading volume; the authors interpret ARIMA as capturing short-term linear dependencies while the LSTM captures nonlinear and long-term structure [2406.18206]. A different hybrid combines ARIMA with a polynomial classifier through
\[
Y_{\text{final}} = \alpha Y_{\text{ARIMA}} + (1-\alpha)Y_{PC},
\]
with \(\alpha\) chosen analytically to minimize MSE [2505.06874].

A still more structural extension is BHT-ARIMA for multiple short time series. There, multi-way delay embedding produces block Hankel tensors, Tucker decomposition compresses them to core tensors, and a tensor-valued ARIMA is applied to the core sequence:
\[
\Delta^d  {\widehat{G}_{t}
=
\sum_{i=1}^{p} \alpha_i  {\Delta^d   \widehat{G}_{t-i}
-
\sum_{i=1}^{q} \beta_i {\widehat{E}_{t-i}
+
\widehat{E}_t.
\]
This preserves the \(p,d,q\) logic while changing the forecasted object from a scalar series to a low-rank tensor state [2002.12135].

## 5. Applications across domains

The papers collectively show that ARIMA and ARIMA-family models are used across epidemiology, climate, finance, astronomy, telecommunications, water systems, insurance, electricity, and macro-financial forecasting. Representative applications are summarized below.

| Domain | Series and model | Representative finding |
|---|---|---|
| Epidemiology | Weekly malaria incidence, ARIMA\((2,1,0)\) | Selected by lowest AIC \(296.80\) after Box–Cox and differencing [1902.07953] |
| Insurance | Quarterly Ghana insurance penetration, ARIMA\((3,1,0)\) | Lowest AIC \(-398.502\); 13-step-ahead forecasts through 2025 Q4 [2502.07841] |
| Finance | Bitcoin daily closing price | Short-term prediction works better in stable subperiods; MSE-optimal model for \(w=9\) was ARIMA\((4,1,1)\) [1904.05315] |
| Astronomy | Minute-scale optical turbulence, ARIMA\((4,0,1)\) | Forecast PDFs for image metrics were “1.5–2 times thinner” than unconditional PDFs [1806.09536] |
| Water systems | CPC pressure in Noja, ARMAX\((4,3)(0,0)\) | Validation MAPE \(=0.728697\%\), RMSE \(=0.304615\) m [2512.09717] |
| Telecommunications | Hourly subscriber throughput | ARIMA reported \(92.67\%\) on subscriber data and \(94.8\%\) on the subscriber testing set, faster than CNN [2404.15095] |

Comparative studies, however, do not point in a single direction. In a monthly-temperature study, a convolution-plus-LSTM model achieved MSE \(1.9006\) versus ARIMA’s \(2.4214\) on the final 12-month test window [2011.04452]. In economics and financial series, a fixed ARIMA\((5,1,0)\) baseline was outperformed by LSTM, with average RMSE reductions of \(87.445\%\) for financial series and \(84.394\%\) for economic series [1803.06386]. Yet for daily cumulative COVID-19 cases, ARIMA\((1,2,2)\) remained broadly comparable to several LSTM variants, with kMAPE as low as \(0.07\%\) and kMdSA as low as \(0.06\%\) [2006.13852]. This suggests that ARIMA’s empirical status is domain- and protocol-dependent rather than uniformly superseded.

## 6. Limitations, misconceptions, and contemporary perspective

A recurrent limitation is linearity. One paper explicitly states that ARIMA is a “linear regression-based forecasting approach” and highlights its difficulty modeling nonlinear relationships [1803.06386]. Financial examples make this concrete. In Bitcoin forecasting, ARIMA performs reasonably well for short-term, one-day prediction in subperiods “in which the behavior of the time-series is almost unchanged,” but introduces large errors over a three-year span with regime change and cannot capture the sharp fluctuations at the end of 2017 [1904.05315].

A second limitation is that differencing alone does not guarantee superiority. In the annual global temperature-anomaly uncertainty study, the original series failed the ADF test, so ARIMA\((1,d,1)\) with \(d=1,2\) was examined, yet the authors concluded that an uncertainty-derived ARMA\((1,1)\) fit better than ARIMA\((1,1,1)\) and ARIMA\((1,2,0)\) on residual diagnostics [2303.02070]. This directly contradicts the misconception that a nonstationary-looking series should automatically lead to an integrated model.

A third limitation is methodological rather than theoretical: ARIMA performance depends heavily on specification, diagnostics, and the fairness of the baseline. The temperature study introduces ARIMA through standard Box–Jenkins steps but does not report the final \((p,d,q)\), provide ACF/PACF plots, or include residual whiteness results; it also omits a seasonal term despite monthly temperature data [2011.04452]. This suggests that some claims of neural superiority over ARIMA may partly reflect under-specified classical baselines rather than an intrinsic inferiority of the methodology.

At the same time, it would be inaccurate to treat ARIMA as merely obsolete. Across the surveyed work, it remains the standard benchmark classical forecaster, often the best single linear model, and frequently the component that supplies interpretable short-memory structure inside hybrids [2010.11617]. A plausible implication is that ARIMA’s contemporary role is dual: it remains a practical end model for short-horizon, univariate, stationarity-adjusted forecasting, and it also functions as a modular building block in exogenous, heteroskedastic, tensor, and neural hybrid systems.

Source: https://www.emergentmind.com/topics/arima