Getting Started

Installation

To install traval, a working version of Python 3.9 or higher has to be installed on your computer.

To install traval, use:

pip install traval

To install in development mode, clone the repository, then type the following from the module root directory:

pip install -e .

Usage

The basic usage of the module is described below. To start using the module, import the package:

import traval

The first step is generally to define an error detection algorithm. This is done with the RuleSet object:

ruleset = traval.RuleSet("my_first_algorithm")

Add a detection rule (using a general rule from the library contained within the module). In this case the rule states any value above 10.0 is suspect:

ruleset.add_rule("rule1",
                 traval.rulelib.rule_ufunc_threshold ,
                 apply_to=0,
                 kwargs={"ufunc": (np.greater,), "threshold": 10.0}
                 )

Take a look at the ruleset by just typing ruleset:

ruleset

RuleSet: 'my_first_algorithm'
   step: name            apply_to
      1: rule1                  0

Next define a Detector object. This object is designed to store a time series and the intermediate and final results after applying an error detection algorithm. Initialize the Detector object with some time series. In this example we assume there is a time series called raw_series:

detect = traval.Detector(raw_series)

Apply our first algorithm to the time series.

detect.apply_ruleset(ruleset)

By default, the result of each step in the algorithm is compared to the original series and stored in the detect.comparisons attribute. Take a look at the comparison between the raw data and the result of the error detection algorithm.

Since we only defined one step, step 1 represents the final result.

cp = detect.comparisons[1]  # result of step 1 = final result

The SeriesComparison* objects contain methods to visualize the comparison, or summarize the number of observations in each category:

cp.plots.plot_series_comparison()  # plot a comparison
cp.summary  # series containing number of observations in each category

For more detailed explanation and more complex examples, see the notebook(s) in the examples directory.