This is useful because these ideas can become input features in a feature engineering effort or simple models that may be combined in an ensembling effort later. This is what makes it so easy to understand and so quick to implement and evaluate.Īs a machine learning practitioner, it can also spark a large number of improvements. It assumes nothing about the specifics of the time series problem to which it is applied. We have seen an example of the persistence model developed from scratch for the Shampoo Sales problem. Test_score = mean_squared_error(test_y, predictions) Series = read_csv('shampoo-sales.csv', header=0, parse_dates=, index_col=0, squeeze=True, date_parser=parser)ĭataframe = concat(, axis=1) The original dataset is credited to Makridakis, Wheelwright, and Hyndman (1998).īelow is a sample of the first 5 rows of data, including the header row.įrom trics import mean_squared_error The units are a sales count and there are 36 observations. This dataset describes the monthly number of shampoo sales over a 3 year period. First, let’s review the Shampoo Sales dataset. To make this concrete, we will look at how to develop a persistence model and use it to establish a baseline performance for a simple univariate time series problem. This satisfies the three above conditions for a baseline forecast. The persistence algorithm uses the value at the previous time step (t-1) to predict the expected outcome at the next time step (t+1). The equivalent technique for use with time series dataset is the persistence algorithm. This could be used for time series, but does not respect the serial correlation structure in time series datasets. This algorithm predicts the majority class in the case of classification, or the average outcome in the case of regression. The most common baseline method for supervised machine learning is the Zero Rule algorithm. Persistence Algorithm (the “naive” forecast) Repeatable: A method that is deterministic, meaning that it produces an expected output given the same input.Ī common algorithm used in establishing a baseline performance is the persistence algorithm.Fast: A method that is fast to implement and computationally trivial to make a prediction.Simple: A method that requires little or no training or intelligence.Three properties of a good technique for making a baseline forecast are: The goal is to get a baseline performance on your time series forecast problem as quickly as possible so that you can get to work better understanding the dataset and developing more advanced models. Once prepared, you then need to select a naive technique that you can use to make a forecast and calculate the baseline performance. The performance measure you intend to use to evaluate forecasts (e.g.The resampling technique you intend to use to estimate the performance of the technique (e.g.The dataset you intend to use to train and evaluate models.The technique used to generate a forecast to calculate the baseline performance must be easy to implement and naive of problem-specific details.īefore you can establish a performance baseline on your forecast problem, you must develop a test harness. If a model achieves performance at or below the baseline, the technique should be fixed or abandoned. It is a point of reference for all other modeling techniques on your problem. Forecast Performance BaselineĪ baseline in forecast performance provides a point of comparison. How to Make Baseline Predictions for Time Series Forecasting with Python
0 Comments
Leave a Reply. |