Questions tagged [forecasting]
Prediction of the future events. It is a special case of [prediction], in the context of [time-series].
3,928 questions
0
votes
0
answers
30
views
Plausibility of results for PCR/PLSR daily stock return forecasting
I'm working on a project for my master's degree and, I am not sure, whether the results I'm getting are plausible or not.
I am basically trying to create a model for forecasting S&P 500 return ...
1
vote
1
answer
32
views
Identifying Poorly Forecastable Time Series Using tsfeatures
I am working on a problem involving the identification of poorly forecastable time series using features extracted with the tsfeatures library by Rob J. Hyndman. Below are the key details about my ...
0
votes
0
answers
24
views
Difficulty in Deriving a Estimator Using Survey Means from Individual Forecasts
I would like to clarify a doubt regarding the paper Testing the Rationality of Price Forecasts: New Evidence from Panel Data (by MICHAEL P. KEANE AND DAVID E. RUNKLE) that presents an estimator ...
- 121
0
votes
0
answers
25
views
AIC from sarima vs arima functions in R
I am doing a report of time series, and while analyzing the time series in R, I noticed the using the sarima function
...
- 51
0
votes
0
answers
7
views
How should I input and output feature and target timeseries to timeseries transformer
I am trying out PatchTST timeseries transformer (paper, code) on a timeseries data that I have. The way PatchTST handles data is as follows:
Note that on line 78-79, the repo does following:
...
- 235
4
votes
1
answer
132
views
How to approach time series forecasting
I am working on a time series forecasting problem involving high-frequency data (hourly or every 10-15 minutes), such as energy consumption or other IoT device metrics. My goal is to predict the ...
- 83
0
votes
0
answers
8
views
Cointegration when one variable is seasonal and another is not
I have two variables: one has a seasonal 12-month pattern and another is seasonally adjusted. I need to make a long-term forecast for 10 years. To do this, one can try out an Error Correction Model if ...
- 1
3
votes
1
answer
52
views
How to best forecast a time series showing level changes and square wave kind of behavior with noise
This is AC power data measured at 1 min interval from March-Dec 2019. I want to model the time series but the out of sample forecast is essentially constant. I found the following from EDA:
Power is ...
- 31
6
votes
1
answer
128
views
Is there a way to forecast by subgroup without forecasting each subgroup separately?
I am trying to find an appropriate model to forecast the number of applications received at the end of a recruitment cycle based on previous recruitment cycles and the number of applications received ...
2
votes
0
answers
16
views
Forecast optimality for categorical dependent variable
I am familiar with several criteria of forecast optimality for variables on a ratio scale. E.g. Diebold Forecasting in Economics, Business, Finance and Beyond introduces the unforecastability ...
- 69.5k
1
vote
0
answers
19
views
ML approaches to pricing in industry?
This is a very vague question. I’m curious how pricing is done in industry, specifically at large tech companies. I know that Amazon does not price discriminate based on user, so price experiments are ...
- 3,520
0
votes
0
answers
18
views
How to Forecast Sales for Sub-Locations Without Historical Proportion Data?
I have a time series dataset of total sales for a product in a store over time. This product is available in two different locations within the store: one stand near the checkout and another stand in ...
0
votes
0
answers
14
views
Under which conditions does PCA consistently estimate latent factors in a Dynamic Factor Model?
Consider a dataset of N time series and T observation periods.
Assume each series $x_t$ is generated from a single (unobserved) common factor $f_t$ following this model:
$$
X_t = \Lambda f_t + \...
0
votes
0
answers
50
views
Making forecasts based on cyclical data
I am trying to develop a more robust methodology for a forecast model. This attempts to project the final number of recruits for this cycle based on comparing the current recruitment cycle to previous ...
2
votes
1
answer
34
views
Fitted values of initial observations in auto.arima for non-stationary models
If I understand correctly, the fitted values returned in the auto.arima of the forecast R package are the one-step ahead forecasts given by the model, once the ...
- 151
3
votes
1
answer
151
views
Cannot reproduce Continuous Ranked Probability Score (CRPS) from Python package
The Continuous Ranked Probability Score (CRPS) is given by:
\begin{equation}
\mathrm{CRPS}(F, x) = \int_{-\infty}^{\infty} \left( F(y) - \mathbb{1}(y - x) \right)^2 \, dy
\end{equation}
I am trying to ...
- 1,244
0
votes
0
answers
7
views
Forecast calibration with a general explanatory factor
I'm thinking of making a set of probability-based prediction for the coming 2 years of the Trump administration. The prediction will be a set of propositions like:
There will be a federal ban on ...
- 146
2
votes
0
answers
12
views
Predicting a jobs cost based on monthly payments (Timeseries forecasting)
Imagine theres a company that hires a cleaning crew each month. Payments are made in the following months, bit by bit.
As depicted below:
In Jan the company paid the cleaning crew ...
huddie96
0
votes
0
answers
14
views
Name for Heidke skill score with model-free reference model
The Heidke skill score is a popular measure for quantifying forecasting skill. It follows the general definition of a skill score (SS):
$$SS = \frac{l_m-l_r}{l_p-l_r},$$
with $l_p$ the loss of a ...
- 1,870
1
vote
0
answers
28
views
Recursive one-step forecasting in timeseries model
I am trying to implement a recursive one-step forecasting approach for a Random Forest model.
The idea is to get a 12-months forecast in an iterative way where each prediction becomes part of the ...
- 140
0
votes
0
answers
17
views
Deriving a multiple based on actuals and forecast values
For context, we are using the DeepAR model for demand planning forecasting. Currently the forecast often underrepresents actual demand. It was suggested that we use a higher quantile to overestimate ...
- 101
0
votes
0
answers
24
views
Forecasting for Multivariate Time Series on Multiple Subject
Let's imagine I have time series data for 50 users and 20 features per user: User1_ts(F1,...F20), User2_ts(F1,...F20), ...User50_ts(F1,...F20). F20 is my target variable, and the goal is to apply to ...
- 131
4
votes
1
answer
236
views
Huge Bounds on Forecast Interval for ARIMA model
I'm working on a financial forecast and trying out a simple SARIMA model to try and forecast the year of 2025. I think the predicted values and fit of the model are fine, and I've tested out many ...
- 43
8
votes
1
answer
290
views
Should out-of-sample validation also be out-of-time for time-series?
Introduction
When training a model a "sample" usually refers to the data used to fit the model, so...
Sample: Data used for training model
Out-of-sample: Data not used for training model
Out-...
- 245
3
votes
1
answer
131
views
Arima function in R incorrectly including an additional MA term
When fitting an ARIMA model (with Arima function from the forecast package) there is an additional hidden MA term.
...
- 33
2
votes
1
answer
34
views
Is it possible to train Neural networks for time series forecasting using elastic distances (such as dtw) as a loss function?
Normally, elastic distances are used as ways to tell how similar two time series are. Examples of these are dynamic time warping and move-split-merge and many more. And I read some researches such as ...
1
vote
1
answer
79
views
Predicting Sales Volume for Complex Customer Base with Time Series Data
I'm working on a time series problem and would appreciate advice from the community. My goal is to predict the sales volume ordered per customer, per product family, for the next 2-4-6 weeks. The data ...
- 11
3
votes
2
answers
79
views
Predicting the Next Event's Timestamp Based on Historical Data with Possible Patterns?
I'm working on a personal project where I aim to predict the time of the next event based on a series of historical timestamps. The dataset I have consists of around 400k timestamps of past events.
...
- 131
1
vote
0
answers
72
views
The Notion of "Predictability" (in Forecasting)
Looking for thought partners to help me clarify a shower thought.
Let's assume I'm a forecaster, and I have in front of me several events with binary outcome–– e.g. a bent coin toss, a two-candidate ...
- 171
1
vote
1
answer
27
views
Is there away to compute Index values (base 100) from Year-over-Year % change (YoY) of the variable?
Let's assume I have a time series like this :
Time period
YoY Change (%)
Y2024 _ Q1
7.00
Y2024 _ Q2
4.85
Y2024 _ Q3
5.77
Y2024 _ Q4
5.66
Y2025 _ Q1
6.54
Y2025 _ Q2
6.48
Y2025 _ Q3
6.36
Y2025 ...
- 109
2
votes
0
answers
27
views
Can we apply Fourier transform on non stationary data?
Hi, I'm trying to predict US inflation rate. The unit is in percentage change from a year ago. Would it be possible to use Fourier transform on the independent data to create a new feature, knowing ...
- 141
1
vote
1
answer
22
views
Decomposing forecast errors into shape and scale components
I'm working on a time series forecasting problem where I predict hourly demand values for each day (24 hours in total). I measure the error of my predictions at the day-level using the Mean Absolute ...
- 111
1
vote
0
answers
23
views
Why is exponential smoothing forecast the median of the forecast density?
I am reading Hyndman & Athanasopoulos "Forecasting: Principles and Practice" 2nd edition (FPP2). (I am aware that 3rd edition exists.) In the chapter about exponential smoothing, section ...
- 69.5k
0
votes
0
answers
23
views
Testing the predictive power of a time series signal
Given a signal from an unknown source/origins and some financial close price data, what are some straightforward ways to test for the signal's predictive power? Would I need to make both series ...
- 1
0
votes
0
answers
16
views
Holt Winters with binary values
I have a question regarding Holt Winters method.
I have an order history which contains at least the customers and the belonging time when the order was taken. We will skip the rest for now since it ...
- 1
1
vote
0
answers
28
views
Predicing next customers purchase dates (and possibly amount)
I have a dataset with simple list of customer, date of purchase, amount. I'd like to predict the next purchase date for each customer and possibly the amount.
Customer
Date of Purchase
Amount
A
05/...
- 11
1
vote
1
answer
47
views
Time-series forecasting problem in Python
I am working on a Python project where I have to predict the energy consumption in individual households. My dataset consists of several thousands of households each having a monthly value of the ...
- 11
2
votes
1
answer
69
views
ARIMA Model Forecasting - 95% Prediction Intervals
I'm currently learning about time series forecasting and ARIMA models. For this question, I'll just be using the AR(1) model example $X_t = \phi X_{t-1} + \varepsilon_t$ and say we are forecasting $X_{...
3
votes
2
answers
36
views
Should Correlation Between Out-of-Sample Forecasts and Actual Values Be Included in Forecast Evaluation Alongside RMSE?
I'm currently evaluating the performance of a forecasting model using the Root Mean Square Error (RMSE) as the primary metric. I've noticed that some literature also includes the Pearson correlation ...
- 363
0
votes
0
answers
33
views
Time series Forecast
I’ve been meaning to ask an expert in the field of forecasting a burning question about a forecasting project am working on. I’m building a model to forecast groundwater thickness, and am tasked to ...
0
votes
0
answers
18
views
Suggestions on what model to use (ARMA doesn't seem to be applicable)
My data looks like the following:
The ACF and PACF plots look like the following:
Although there is some dependence with some lags, I fear taking these too seriously is a form of overfitting. I don'...
- 61
1
vote
1
answer
58
views
Why do the simulations of my SARIMA model not resemble my original data?
I want to simulate a SARIMA model I obtained using the auto.arima function from the R package "forecast". My objective is to be able to do a lot of simulations in order to "predict"...
- 11
0
votes
0
answers
23
views
New to Forecasting -- Any tips for Proper Model to Use?
So I've done a ton of internet research and lost. I am trying to forecast dayrates for offshore drillships. and I have two exogenous variables that have close correlation with my historical dayrate ...
0
votes
0
answers
42
views
CausalImpact package in R - One-tailed probability
This is my first time using Bayesian statistics and the CausalImpact package in R. I'm a bit confused about whether this is using one-tail or two-tailed probability testing and was wondering if anyone ...
1
vote
1
answer
45
views
Why ARIMA on same data says data is white noise for one forecast duration while not indicating that it's not white noise on a different duration
I am running a ARIMA model on my data. I have weekly data from Jan 2021. When I run 12 weeks forecast, the ARIMA gives the best parameter values (0,0,0) indicating that the data is white noise. But ...
- 73
0
votes
0
answers
15
views
Direct Multi-Step Forecasting by each day of the week?
I am looking to create a daily 7 day forecast for a particular domain problem.
Currently I have a recursive solution, which hasn't been performing too well.
I have also looked into direct forecasting ...
- 23
2
votes
1
answer
55
views
How to validate the predictions from the function forecast in the R?
Motivation
We are often interested in evaluating a model's performance when it encounters previously unseen data. I am testing ARIMA models to analyze some data and want to ensure I understand the ...
- 706
1
vote
0
answers
30
views
Using monthly data for yearly predictions
I need to predict a yearly quantity $y$ (water flowing into a reservoir). This is based on several factors, e.g. $x$ (amount of rain in the current year), and probably last years quantities (e.g., $...
- 3,680
0
votes
0
answers
22
views
How do I quantify the significance of a percentage ranked data set?
I'm considering seven different forecast models for a given data set, over a number of forecast iterations, and quantifying the number of times each forecast model was the most accurate overall.
I ...
1
vote
0
answers
67
views
Prediction of a time series AR(1) vs AR(1) with exogenous variables vs Random Forest, why is the performance so different?
Extension of the previous question that only compare AR(1) vs. AR(1) with exogenous variables:
I am currently working on forecasting a time series y using three models: an AR(1) model and an AR(1) ...
- 363