Exact Discovery of Time Series Motifs
Discover recurring patterns (motifs) in 1-D time series using a modified version of the algorithm described in Mueen et al., EMMA. It uses Pandas and returns motifs as a structured DataFrame.
- Detect motifs in a 1-D time series (
pandas.Series). - Cross-platform support: Windows, Linux, macOS.
- Handles permissions and prompts installation for
wineon non-Windows systems. - Returns motifs in a hierarchical DataFrame with groupings and motif indices.
| Parameter | Type | Description |
|---|---|---|
data |
pd.Series |
One-dimensional time series input. |
motif_length |
int |
Length of each motif subsequence. |
major_factor |
float |
Major factor of cluster radius (>1 and >minor_factor). |
minor_factor |
float |
Minor factor of cluster radius (>1). |
| Output | Type | Description |
|---|---|---|
motifs |
pd.DataFrame |
DataFrame containing all discovered motifs. Columns are hierarchical: Group, Number, Id, and value. |
bsfb |
float |
Final Euclidean distance (best-so-far break) computed by the algorithm. |
DataFrame structure example:
Group | Number | Id | value
------|--------|----|------
0 | 0 | 0 | 1.23
0 | 0 | 1 | 1.45
0 | 1 | 0 | 2.01
...