kevsatt / python-timeseries-segmenter

Simple data segmenter

Install

pip install git+https://github.com/XavierTolza/python-timeseries-segmenter.git

Or add git+https://github.com/XavierTolza/python-timeseries-segmenter.git to you requirements.txt:

Usage

Inputs are pandas dataframe with x axis as index. A plot will be drawn for each column and you need to specify the class with a column labelled "class" See the example below:

# Generate sample data
n_values = 1000
n_classes = 4
columns = "A,B,C,D".split(",")
data = np.cumsum(np.random.normal(0, .1, (n_values, len(columns))), 0)
data = DataFrame(data,
                 index=np.linspace(0, 5, n_values), columns=columns)
# Now data is correct, generate class for each sample
classes = np.array(["Class %i" % i for i in range(n_classes)])[np.linspace(0, n_classes-1, n_values).astype(int)]
data["class"] = classes

# Run the UI
s = dfs.Segmenter(data)
res = s.run()

# The result is the same dataframe as input, but with classes updated
print(res)

Dependencies

• matplotlib
• numpy
• Does not require pandas, you might have some trouble if you don't have it ;)