Questions tagged [forecasting]
Forecasting is the process predicting future values based on historic and current data, typically for time-series datasets.
337
questions
0
votes
0
answers
3
views
What are some good methods for evaluating the disaggregation ratios in a top down approach?
Say we are forecasting at a high level (Department - Week) and we want to break it down to (Category - Week) level.
I want to find out which department's disaggregation ratios needs to be improved.
0
votes
0
answers
15
views
Building recursive forecasting from scratch?
I am currently testing time-series modelling with tree-based (xgb, lgbm, cat) algorithms using, recursive lags of a y (sales) along with with numerous exogenous regressors i.e. date-time, price, ...
- 101
0
votes
0
answers
7
views
LSTM input_shape returns value error
My time series dataset dimension are as follows:
print(X_train.shape) = (1766, 4) i.e. 1,766 time steps and 4 features
...
0
votes
1
answer
13
views
How to predict probability of an event when we have a month to month data?
I'm trying to find references about how to proceed to get the probability of an event happening when we have "temporal data" in our table
My data is basically:
hex_id: id of the object
date:...
0
votes
0
answers
21
views
Model for predicting temperature data of fridge
I set up a sensor which measures temperature data every 3 seconds. I collected the data for 3 days and have 60.000 rows in my csv export. Now I would like to forecast the next few days. When looking ...
- 1
0
votes
1
answer
47
views
How to train machine learning on sales forecasting problems of almost 10,000 shops?
I have a dataset of almost 10,000 shops, 'dates', 'shop ID' and 'sales amounts' as their features almost 2 years of data. I want to forecast each shop, the sales amount for 30 next days. I want to ...
- 123
2
votes
1
answer
22
views
How does the Kalman Filter actually work?
I know that the Kalman filter can be used whenever we have a measurement and transition equation. I also know that the Kalman filter can handle missing data. From my course at university, I know that ...
- 23
0
votes
1
answer
18
views
Force predictions of 2 time series models with different time steps to be consistent
Suppose I have a time series. Let's say it is of the number of sales in a shop. Suppose I am looking to make two models - model 1 which predicts future values by weekly time steps (total sales per ...
- 1
0
votes
0
answers
12
views
GluonTS with small datasets
I am working with a small dataset which is approx. 60 observations. Is it possible to build models from GluonTS which provides probabilistic forecasting also which makes forecasting more reliable but ...
0
votes
1
answer
26
views
Time-series feature enrichment before or after train-test split?
I am dealing with a time-series that represents the CPU usage registered on Azure Virtual Machine. The historical data include a period of 19 months and its granularity is a 10 minute one (each 10 ...
- 1
0
votes
0
answers
30
views
In ensembles combining models, does it make sense for a model to have negative weight?
I have 13 models ranging from simple models like Seasonal Naïve Average to complex models like Random Forests, The weights of the models is calculated based on the LPMinimize of the error during the ...
- 103
1
vote
1
answer
59
views
How to use 5 different datasets to re-train & test your ML Classification models multiple times
My R scripts and my 5 source datasets can be found in my GitHub Repository for this project, and I originally found this source data on Kaggle. This set of source data includes 5 datasets with over ...
- 239
0
votes
1
answer
19
views
How to forecast time bound contract based revenue?
I'm looking at ideas to see how I can forecast contract-based revenue. For example, I can have customers who have purchased a monthly mobile plan, an annual plan, or a 3-year plan. How can I use that ...
0
votes
0
answers
15
views
Deep learning models for time series forecasting changing the input of the experiment
For my project I received the following pandas dataframe which contains the dissipated heat from an experiment. Each column represents the power provided to the experiment. The distance between each ...
- 1
0
votes
0
answers
11
views
Do we need padding in global time series models?
In order to forecast product demand for a costumer, I need to predict time series of different lengths.
I am using mostly global lightgbm models to do the job.
In most use cases, it is common practice,...
- 145
2
votes
1
answer
77
views
How to use categorical data in forecasting with Prophet?
I'm trying to create a model to predict the number of players on a video game at a certain time and was trying to figure out how to integrate categorical data into my forecasting problem. So far, my ...
0
votes
0
answers
68
views
TimeSeries forecasting with Catboost
After extensive research on both the documentation and internet itself, I found many articles showing how to fit() and predict() a CatboostRegressor, but all of them use split data for train/test (...
1
vote
0
answers
13
views
Time series forecasting with incomplete future data
I have historical daily time series bookings data. I am using the Prophet model to predict daily bookings for the next 7 days. However, I also have incomplete booking information for future dates.
...
- 155
0
votes
0
answers
12
views
How to aggregate effects of time series, VAR and linear regression on the same dataset?
I have the Walmart store data from here
https://www.kaggle.com/competitions/walmart-recruiting-store-sales-forecasting/data
Say I aggregated the data at date level and now want to predict sales.
There ...
- 1
0
votes
2
answers
21
views
What model to fit to call center data
I have a dataset with calls from day 1 to day 340. What model can I fit to mathematically capture the pattern?
There are only 1 or 2 digit number of calls on all days except day 61.62.63 and 121.122....
1
vote
1
answer
72
views
Why so discrepancy between ARIMA and LSTM in time series forecasting?
I have this time series below, that I divided into train, val and test:
Basically, I trained an ARIMA and an LSTM on those data, and results are completely different, in terms of prediction:
ARIMA:
...
- 209
0
votes
0
answers
12
views
Sales forecasting with hyper-parameter optimization for categories
Example
Item Category X (e.g. short sleeve t-shirt) contains two items.
Item A (e.g. short sleeve t-shirt, white)
Item B (e.g. short sleeve t-shirt, red)
Question
Is it logical to do the method below? ...
- 101
0
votes
0
answers
50
views
How to make ARIMA model out of sample forecasts with exog Fourier terms using weekly data? (Python)
I'm a bit confused on how to make out-of-sample predictions if I have Fourier terms included in my ARIMA model. I am using Fourier terms to model annual seasonality as per the advice given in ...
- 1
1
vote
0
answers
30
views
"Most forecasting algorithms assumes that each point is independent of one another." If so, how forecasting is being possible?
In stationary, if we want to make forecasting, we have to make our data stationary(On classical methods), I get that, but If every data point is independent of each other, how can we make the ...
- 53
0
votes
0
answers
22
views
How can I create a prediction model with different input variables
I have the following structure for training data:
Input 1
Input 2
Input 3
Input 4
Input 5
Output 1
0.6
0.8
0.9
0.6
0.8
0.7
0.9
0.4
0.4
0.6
nil
0.6
0.8
0.7
0.7
0.8
nil
0.8
0.7
0.8
0.5
nil
nil
0.5
...
0
votes
1
answer
31
views
Determining 'Addictive' or 'multiplicative' seasonality and its forecast accuracy
Let's say that I have a "train" and "test" set data, how do I determine if my train set follows "additive" or "multiplicative" seasonality? Do I fit just the ...
0
votes
0
answers
36
views
How to select the best 30 features from 500 features for sales prediction model where feature importance can change over time?
I'm using data sets for sales prediction model which is trained every 2 weeks. It
has 200 features and 500 rows. I have to select the best 30 features that can be used in the
model instead of 200 ...
- 101
0
votes
0
answers
11
views
Multistep horizon strategies vs Seq2Seq
In the context of recurrent neural networks for forecasting, what is the relation between multi-step horizon strategies and RNN-based seq2seq?
For example, is the following Multiple Output Strategy [1,...
- 103
0
votes
1
answer
146
views
How does one perform a Canova-Hansen test in Python?
I am referring to the documentation here, but it does not give many examples on how to actually perform the test. I have a pandas dataframe with two columns:
Column 1 is first day of every week,
...
- 5
1
vote
1
answer
31
views
predictions based on irregular repeated measures?
I need to make a model that predicts certain medical outcomes based on the answer to health-related questionnaires. Providers have patients fill out these questionnaires more than once, at irregular ...
- 11
1
vote
0
answers
27
views
Why is Orange (CSV import or Line Series) doing some weird rounding on my data?
The data I have is in tab-separated format (exported from MetaTrader5):
...
0
votes
1
answer
28
views
Detrend a time series
I am fairly new to forecasting and I am trying to create a demand forecast for my organization; I am following the methodology outlined here. In step 12 of the process, the author subtracts the trend ...
- 5
0
votes
0
answers
13
views
Use two different but correlated time-series signals as two different samples to train a model
I want to train a forecasting model for signal A initially. In the future, forecasts B and C may be required. These are all financial time-series signals with the same resolution. The issue I am ...
0
votes
0
answers
9
views
Forecasting physical performance of young football players
I have a dataset containing physical testing data from a football academy, including sprint tests, COD test, endurance tests, strengths tests, and anthropometric features. The dataset contains all ...
0
votes
0
answers
31
views
Why is my LSTM prediction is saturated and have bad prediction?
I am new to deep learning. Currently, I am trying to predict torque based on its past values using an LSTM model. There are two datasets (generated from a scaled test), one with wear and the second ...
- 1
0
votes
0
answers
7
views
is there a standard metric to check volatility between every week forecasting results?
I have fitted arima, tbats model on my dataset and forecasted results for 12weeks. my forecasting results from 1st week to 2nd week differs a lot. I checked how much is the variance(volatility) ...
- 1
0
votes
1
answer
50
views
Time Series Forecasting in Python Question
I am trying to understand what is the best possible way to execute a time series forecast. I am trying to forecast the number of employees that are going to call in sick on a given day. My data has ...
- 11
0
votes
0
answers
27
views
Time Series Forecast: how to deal with NA-values in lagged predictors of the target variable?
For my time series forecast of several consumer products, I am using lags from the target variable.
In order to forecast 24 month, I am using 24 lags (24 tend to be a bit more accurate than 12, but ...
- 145
0
votes
0
answers
8
views
Which time series model to choose?
I am new to time series forecasting. I have two very large dataset consisting of about 530k values obtained from a scaled dataset. The nature of both of these datasets are different. One of them has ...
- 1
1
vote
1
answer
230
views
Sktime - how to make in-sample and out-of-sample forecasts with exogenous variables?
I'm trying to make forecasts using sktime for my entire training data and an arbitrary length of out-of-sample data but can't figure it out.
...
- 226
0
votes
1
answer
16
views
How do I prepare time-series data to predict the change in a response variable, rather than the value of the response variable itself?
I am forecasting solar irradiance using different time series models. Rather than predict the irradiance from $t=0 \text{ min}$ to $t=120 \text{ min}$, I would like to predict the change in irradiance ...
- 101
2
votes
1
answer
70
views
ARIMA predictions look shifted by one unit of time
I am using statsmodels ARIMA (1,2,1) to predict the monthly demand for a product. The predictions look like they are shifted to the right by one month. I wonder if the statsmodels.ARIMA.Residuals....
- 121
1
vote
0
answers
23
views
Demand Forecasting/Regression task for new products
I'm currently at the end of my master's degree and have to solve a data science problem. I am currently kind of stuck and need some kind of advice to get better results. I want to share the task I ...
0
votes
0
answers
36
views
Walk-forward validation nrounds LightGBM
Dear Data Science Gurus,
I am facing the following phenomenon:
1-Scenario:
Predict demand for n-month in the future for hundreds of products using lightgbm. I have 2 years of history
2- Approach:
Walk-...
0
votes
0
answers
18
views
Time series forecasting model predicts increasing number for target variable when the actual values are zeroes
I am working on time series forecasting model, and I am using light gbm. The project goal is to predict the number of sales across different levels (very similar to the M5 competition). For instance, ...
- 113
0
votes
0
answers
42
views
0
votes
2
answers
53
views
Anomaly prediction/forecasting in timeseries?
What options exist in order to forecast when next observation will be an outlier in a time series? Initially, I thought to train a simple forecasting model, which turned out to decently predict the ...
- 1
0
votes
1
answer
41
views
0
votes
0
answers
23
views
What model to use to forecast on small dataset with many parameters?
I have a dataset of 20 columns/parameters (x's) and not many rows (10-20 historical values for each x) I need to use to predict a y column for the future 5 years. Each row represents one year. What ...
- 1
1
vote
0
answers
73
views
predict next purchase time of an item
I have a bunch of timestamps (purchase date from history), that looks like:
[1658753101, 1658760061, 1658824861, 1658846461, 1658853961, etc]
What I want is to based on that list predict next item ...
- 11