PyTrendy is a robust solution for identifying and analyzing trends in time series. Unlike other trend detection packages, it is robust to noisy & flat segments, and handles for gradual & abrupt trend cases with a high precision. It aims to be the best package for trend detection in python.
Install the package from PyPi.
pip install pytrendy
Import pytrendy.
import pytrendy as ptLoad daily time series data. In this case, we're using one of pytrendy's custom examples.
df = pt.load_data('series_synthetic')
print(df)
# date abrupt gradual gradual-noisy-20
# 0 2025-01-01 19.578066 12.500000 27.514106
# 1 2025-01-02 19.358378 13.421717 -6.620099
# 2 2025-01-03 19.228408 13.474026 22.122134
# 3 2025-01-04 19.727130 13.474026 13.863735
# 4 2025-01-05 20.773716 14.505772 8.884535
# .. ... ... ... ...
# 176 2025-06-26 4.718725 20.616883 19.790026
# 177 2025-06-27 4.242065 20.978084 19.181404
# 178 2025-06-28 6.012296 22.449495 -6.563936
# 179 2025-06-29 4.603068 23.486652 48.291088
# 180 2025-06-30 4.435105 22.240260 3.343233Run trend detection & plot the results.
results = pt.detect_trends(df, date_col='date', value_col='gradual', plot=True)
The results object can be used to summarise, further analyse, and generally inspect the trend detections.
results.print_summary()
# Detected:
# - 3 Uptrends.
# - 3 Downtrends.
# - 3 Flats.
# - 0 Noise.
# The best detected trend is Down between dates 2025-05-09 - 2025-06-17
# Full Results:
# -------------------------------------------------------------------------------
# direction start end days total_change change_rank
# time_index
# 1 Up 2025-01-02 2025-01-24 22 14.013348 5
# 2 Down 2025-01-25 2025-02-05 11 -13.564214 6
# 3 Flat 2025-02-06 2025-02-09 3 NaN 7
# 4 Up 2025-02-10 2025-03-14 32 24.632035 3
# 5 Flat 2025-03-15 2025-03-17 2 NaN 8
# 6 Down 2025-03-18 2025-04-01 14 -22.721861 4
# 7 Up 2025-04-02 2025-05-08 36 72.611833 2
# 8 Down 2025-05-09 2025-06-17 39 -73.253968 1
# 9 Flat 2025-06-18 2025-06-30 12 NaN 9
# -------------------------------------------------------------------------------You can directly call the object as a pandas dataframe. Note change_rank which prioritises long duration and high magnitude of change.
results.df| time_index | direction | start | end | trend_class | change | pct_change | days | total_change | SNR | change_rank |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Up | 2025-01-02 | 2025-01-24 | gradual | 14.013348 | 1.044080 | 22 | 14.013348 | 22.207980 | 5 |
| 2 | Down | 2025-01-25 | 2025-02-05 | gradual | -13.564214 | -0.554982 | 11 | -13.564214 | 17.360657 | 6 |
| 3 | Flat | 2025-02-06 | 2025-02-09 | NaN | NaN | NaN | 3 | NaN | 20.126008 | 7 |
| 4 | Up | 2025-02-10 | 2025-03-14 | gradual | 26.015512 | 1.974942 | 32 | 24.632035 | 18.871430 | 3 |
| 5 | Flat | 2025-03-15 | 2025-03-17 | NaN | NaN | NaN | 2 | NaN | 17.350339 | 8 |
| 6 | Down | 2025-03-18 | 2025-04-01 | gradual | -22.721861 | -0.591909 | 14 | -22.721861 | 16.762790 | 4 |
| 7 | Up | 2025-04-02 | 2025-05-08 | gradual | 73.687771 | 3.944243 | 36 | 72.611833 | 21.701162 | 2 |
| 8 | Down | 2025-05-09 | 2025-06-17 | gradual | -73.253968 | -0.805442 | 39 | -73.253968 | 21.122099 | 1 |
| 9 | Flat | 2025-06-18 | 2025-06-30 | NaN | NaN | NaN | 12 | NaN | 19.418124 | 9 |
- Full documentation with all features [WIP].
- Automated testing in CI/CD pipeline with full code coverage.
- Even more robust edge case testing & generalising.
- Customising more options for windows.