A data preprocessing package for time series data. Design for machine learning and deep learning.

Related tags

Machine Learningdata-sciencedeep-learningtime-seriesstock
Overview

Time Series Transformer

Documentation https://allen-chiang.github.io/Time-Series-Transformer/

made-with-python Build Build Status Board Status CodeFactor

import pandas as pd
import numpy as np
from time_series_transform.sklearn import *
import time_series_transform as tst

Introduction

This package provides tools for time series data preprocessing. There are two main components inside the package: Time_Series_Transformer and Stock_Transformer. Time_Series_Transformer is a general class for all type of time series data, while Stock_Transformer is a sub-class of Time_Series_Transformer. Time_Series_Transformer has different functions for data manipulation, io transformation, and making simple plots. This tutorial will take a quick look at the functions for data manipulation and basic io. For the plot functions, there will be other tutorial to explain.

Time_Series_Transformer

Since all the time series data having time data, Time_Series_Transformer is required to specify time index. The basic time series data is time series data with no special category. However, there a lot of cases that a time series data is associating with categories. For example, inventory data is usually associate with product name or stores, or stock data is having different ticker names or brokers. To address this question, Time_Series_Transformer can specify the main category index. Given the main category index, the data can be manipulated in parallel corresponding to its category.

Here is a simple example to create a Time_Series_Transformer without specifying its category.

data = {
    'time':[1,2,3,4,5],
    'data1':[1,2,3,4,5],
    'data2':[6,7,8,9,10]
}
trans = tst.Time_Series_Transformer(data,timeSeriesCol='time')
trans
data column
-----------
time
data1
data2
time length: 5
category: None

There are two ways to manipulate the data. The first way is use the pre-made functions, and the second way is to use the transform function and provide your custom function. There are six pre-made functions including make_lag, make_lead, make_lag_sequence, make_lead_sequence, and make_stack_sequence. In the following demonstration, we will show each of the pre-made functions.

Pre-made functions

make_lag and make_lead functions are going to create lag/lead data for input columns. This type of manipulation could be useful for machine learning.

trans = tst.Time_Series_Transformer(data,timeSeriesCol='time')
trans = trans.make_lag(
    inputLabels = ['data1','data2'],
    lagNum = 1,
    suffix = '_lag_',
    fillMissing = np.nan
            )
print(trans.to_pandas())
   time  data1  data2  data1_lag_1  data2_lag_1
0     1      1      6          NaN          NaN
1     2      2      7          1.0          6.0
2     3      3      8          2.0          7.0
3     4      4      9          3.0          8.0
4     5      5     10          4.0          9.0

trans = tst.Time_Series_Transformer(data,timeSeriesCol='time')
trans = trans.make_lead(
    inputLabels = ['data1','data2'],
    leadNum = 1,
    suffix = '_lead_',
    fillMissing = np.nan
            )
print(trans.to_pandas())
   time  data1  data2  data1_lead_1  data2_lead_1
0     1      1      6           2.0           7.0
1     2      2      7           3.0           8.0
2     3      3      8           4.0           9.0
3     4      4      9           5.0          10.0
4     5      5     10           NaN           NaN

make_lag_sequence and make_lead_sequence is to create a sequence for a given window length and lag or lead number. This manipulation could be useful for Deep learning.

trans = tst.Time_Series_Transformer(data,timeSeriesCol='time')
trans = trans.make_lag_sequence(
    inputLabels = ['data1','data2'],
    windowSize = 2,
    lagNum =1,
    suffix = '_lag_seq_'
)
print(trans.to_pandas())
   time  data1  data2 data1_lag_seq_2 data2_lag_seq_2
0     1      1      6      [nan, nan]      [nan, nan]
1     2      2      7      [nan, 1.0]      [nan, 6.0]
2     3      3      8      [1.0, 2.0]      [6.0, 7.0]
3     4      4      9      [2.0, 3.0]      [7.0, 8.0]
4     5      5     10      [3.0, 4.0]      [8.0, 9.0]

trans = tst.Time_Series_Transformer(data,timeSeriesCol='time')
trans = trans.make_lead_sequence(
    inputLabels = ['data1','data2'],
    windowSize = 2,
    leadNum =1,
    suffix = '_lead_seq_'
)
print(trans.to_pandas())
   time  data1  data2 data1_lead_seq_2 data2_lead_seq_2
0     1      1      6       [2.0, 3.0]       [7.0, 8.0]
1     2      2      7       [3.0, 4.0]       [8.0, 9.0]
2     3      3      8       [4.0, 5.0]      [9.0, 10.0]
3     4      4      9       [nan, nan]       [nan, nan]
4     5      5     10       [nan, nan]       [nan, nan]

Custom Functions

To use the transform function, you have to create your custom functions. The input data will be passed as dict of list, and the output data should be either pandas DataFrame, pandas Series, numpy ndArray or list. Note, the output length should be in consist with the orignal data length.

For exmaple, this function takes input dictionary data and sum them up. The final output is a list.

import copy
def list_output (dataDict):
    res = []
    for i in dataDict:
        if len(res) == 0:
            res = copy.deepcopy(dataDict[i])
            continue
        for ix,v in enumerate(dataDict[i]):
            res[ix] += v
    return res
trans = tst.Time_Series_Transformer(data,timeSeriesCol='time')
trans = trans.transform(
    inputLabels = ['data1','data2'],
    newName = 'sumCol',
    func = list_output
)
print(trans.to_pandas())
   time  data1  data2  sumCol
0     1      1      6       7
1     2      2      7       9
2     3      3      8      11
3     4      4      9      13
4     5      5     10      15

The following example will output as pandas DataFrame and also takes additional parameters. Note: since pandas already has column name, the new name will automatically beocme suffix.

def pandas_output(dataDict, pandasColName):
    res = []
    for i in dataDict:
        if len(res) == 0:
            res = copy.deepcopy(dataDict[i])
            continue
        for ix,v in enumerate(dataDict[i]):
            res[ix] += v
    return pd.DataFrame({pandasColName:res})
trans = tst.Time_Series_Transformer(data,timeSeriesCol='time')
trans = trans.transform(
    inputLabels = ['data1','data2'],
    newName = 'sumCol',
    func = pandas_output,
    pandasColName = "pandasName"
)
print(trans.to_pandas())
   time  data1  data2  sumCol_pandasName
0     1      1      6                  7
1     2      2      7                  9
2     3      3      8                 11
3     4      4      9                 13
4     5      5     10                 15

Data with Category

Since time series data could be associated with different category, Time_Series_Transformer can specify the mainCategoryCol parameter to point out the main category. This class only provide one columns for main category because multiple dimensions can be aggregated into a new column as main category.

The following example has one category with two type a and b. Each of them has some overlaped and different timestamp.

data = {
    "time":[1,2,3,4,5,1,3,4,5],
    'data':[1,2,3,4,5,1,2,3,4],
    "category":['a','a','a','a','a','b','b','b','b']
}
trans = tst.Time_Series_Transformer(data,'time','category')
trans
data column
-----------
time
data
time length: 5
category: a

data column
-----------
time
data
time length: 4
category: b

main category column: category

Since we specify the main category column, data manipulation functions can use n_jobs to execute the function in parallel. The parallel execution is with joblib implmentation (https://joblib.readthedocs.io/en/latest/).

trans = trans.make_lag(
    inputLabels = ['data'],
    lagNum = 1,
    suffix = '_lag_',
    fillMissing = np.nan,
    n_jobs = 2,
    verbose = 10        
)
print(trans.to_pandas())
[Parallel(n_jobs=2)]: Using backend LokyBackend with 2 concurrent workers.


   time  data  data_lag_1 category
0     1     1         NaN        a
1     2     2         1.0        a
2     3     3         2.0        a
3     4     4         3.0        a
4     5     5         4.0        a
5     1     1         NaN        b
6     3     2         1.0        b
7     4     3         2.0        b
8     5     4         3.0        b


[Parallel(n_jobs=2)]: Done   2 out of   2 | elapsed:    3.6s remaining:    0.0s
[Parallel(n_jobs=2)]: Done   2 out of   2 | elapsed:    3.6s finished

To further support the category, there are two functions to deal with different time length data: pad_different_category_time and remove_different_category_time. The first function is padding the different length into same length, while the other is remove different timestamp.

trans = tst.Time_Series_Transformer(data,'time','category')
trans = trans.pad_different_category_time(fillMissing = np.nan
)
print(trans.to_pandas())
   time  data category
0     1   1.0        a
1     2   2.0        a
2     3   3.0        a
3     4   4.0        a
4     5   5.0        a
5     1   1.0        b
6     2   NaN        b
7     3   2.0        b
8     4   3.0        b
9     5   4.0        b

trans = tst.Time_Series_Transformer(data,'time','category')
trans = trans.remove_different_category_time()
print(trans.to_pandas())
   time  data category
0     1     1        a
1     3     3        a
2     4     4        a
3     5     5        a
4     1     1        b
5     3     2        b
6     4     3        b
7     5     4        b

IO

IO is a huge component for this package. The current version support pandas DataFrame, numpy ndArray, Apache Arrow Table, Apache Feather, and Apache Parquet. All those io can specify whether to expand category or time for the export format. In this demo, we will show numpy and pandas. Also, Transformer can combine make_label function and sepLabel parameter inside of export to seperate data and label.

pandas

data = {
    "time":[1,2,3,4,5,1,3,4,5],
    'data':[1,2,3,4,5,1,2,3,4],
    "category":['a','a','a','a','a','b','b','b','b']
}
df = pd.DataFrame(data)
trans = tst.Time_Series_Transformer.from_pandas(
    pandasFrame = df,
    timeSeriesCol = 'time',
    mainCategoryCol= 'category'
)
trans
data column
-----------
time
data
time length: 5
category: a

data column
-----------
time
data
time length: 4
category: b

main category column: category

To expand the data, all category should be in consist. Besides the pad and remove function, we can use preprocessType parameter to achive that.

print(trans.to_pandas(
    expandCategory = True,
    expandTime = False,
    preprocessType = 'pad'
))
   time  data_a  data_b
0     1       1     1.0
1     2       2     NaN
2     3       3     2.0
3     4       4     3.0
4     5       5     4.0

print(trans.to_pandas(
    expandCategory = False,
    expandTime = True,
    preprocessType = 'pad'
))
   data_1  data_2  data_3  data_4  data_5 category
0       1     2.0       3       4       5        a
1       1     NaN       2       3       4        b

print(trans.to_pandas(
    expandCategory = True,
    expandTime = True,
    preprocessType = 'pad'
))
   data_a_1  data_b_1  data_a_2  data_b_2  data_a_3  data_b_3  data_a_4  \
0         1       1.0         2       NaN         3       2.0         4   

   data_b_4  data_a_5  data_b_5  
0       3.0         5       4.0

make_label function can be used with sepLabel parameter. This function can be used for seperating X and y for machine learning cases.

trans = trans.make_lead('data',leadNum = 1,suffix = '_lead_')
trans = trans.make_label("data_lead_1")
data, label = trans.to_pandas(
    expandCategory = False,
    expandTime = False,
    preprocessType = 'pad',
    sepLabel = True
)
print(data)
   time  data category
0     1   1.0        a
1     2   2.0        a
2     3   3.0        a
3     4   4.0        a
4     5   5.0        a
5     1   1.0        b
6     2   NaN        b
7     3   2.0        b
8     4   3.0        b
9     5   4.0        b

print(label)
   data_lead_1
0          2.0
1          3.0
2          4.0
3          5.0
4          NaN
5          2.0
6          NaN
7          3.0
8          4.0
9          NaN

numpy

Since numpy has no column name, it has to use index number to specify column.

data = {
    "time":[1,2,3,4,5,1,3,4,5],
    'data':[1,2,3,4,5,1,2,3,4],
    "category":['a','a','a','a','a','b','b','b','b']
}
npArray = pd.DataFrame(data).values
trans = tst.Time_Series_Transformer.from_numpy(
    numpyData= npArray,
    timeSeriesCol = 0,
    mainCategoryCol = 2)
trans
data column
-----------
0
1
time length: 5
category: a

data column
-----------
0
1
time length: 4
category: b

main category column: 2

trans = trans.make_lead(1,leadNum = 1,suffix = '_lead_')
trans = trans.make_label("1_lead_1")
X,y = trans.to_pandas(
    expandCategory = False,
    expandTime = False,
    preprocessType = 'pad',
    sepLabel = True
)
print(X)
   0    1  2
0  1  1.0  a
1  2  2.0  a
2  3  3.0  a
3  4  4.0  a
4  5  5.0  a
5  1  1.0  b
6  2  NaN  b
7  3  2.0  b
8  4  3.0  b
9  5  4.0  b

print(y)
   1_lead_1
0       2.0
1       3.0
2       4.0
3       5.0
4       NaN
5       2.0
6       NaN
7       3.0
8       4.0
9       NaN

Stock_Transformer

Stock_Transformer is a subclass of Time_Series_Transformer. Hence, all the function demonstrated in Time_Series_Transformer canbe used in Stock_Transformer. The differences for Stock_Transformer is that it is required to specify High, Low, Open, Close, Volume columns. Besides these information, it has pandas-ta strategy implmentation to create technical indicator (https://github.com/twopirllc/pandas-ta). Moreover, the io class for Stock_Transformer support yfinance and investpy. We can directly extract data from these api.

create technical indicator

stock = tst.Stock_Transformer.from_stock_engine_period(
    symbols = 'GOOGL',period ='1y', engine = 'yahoo'
)
stock
data column
-----------
Date
Open
High
Low
Close
Volume
Dividends
Stock Splits
time length: 253
category: None

import pandas_ta as ta
MyStrategy = ta.Strategy(
    name="DCSMA10",
    ta=[
        {"kind": "ohlc4"},
        {"kind": "sma", "length": 10},
        {"kind": "donchian", "lower_length": 10, "upper_length": 15},
        {"kind": "ema", "close": "OHLC4", "length": 10, "suffix": "OHLC4"},
    ]
)
stock = stock.get_technial_indicator(MyStrategy)
print(stock.to_pandas().head())
         Date         Open         High          Low        Close   Volume  \
0  2020-01-06  1351.630005  1398.319946  1351.000000  1397.810059  2338400   
1  2020-01-07  1400.459961  1403.500000  1391.560059  1395.109985  1716500   
2  2020-01-08  1394.819946  1411.849976  1392.630005  1405.040039  1765700   
3  2020-01-09  1421.930054  1428.680054  1410.209961  1419.790039  1660000   
4  2020-01-10  1429.469971  1434.939941  1419.599976  1428.959961  1312900   

   Dividends  Stock Splits        OHLC4  SMA_10  DCL_10_15  DCM_10_15  \
0          0             0  1374.690002     NaN        NaN        NaN   
1          0             0  1397.657501     NaN        NaN        NaN   
2          0             0  1401.084991     NaN        NaN        NaN   
3          0             0  1420.152527     NaN        NaN        NaN   
4          0             0  1428.242462     NaN        NaN        NaN   

   DCU_10_15  EMA_10_OHLC4  
0        NaN           NaN  
1        NaN           NaN  
2        NaN           NaN  
3        NaN           NaN  
4        NaN           NaN

For more usage please visit our gallery

You might also like...
Automatically build ARIMA, SARIMAX, VAR, FB Prophet and XGBoost Models on Time Series data sets with a Single Line of Code. Now updated with Dask to handle millions of rows.
Automatically build ARIMA, SARIMAX, VAR, FB Prophet and XGBoost Models on Time Series data sets with a Single Line of Code. Now updated with Dask to handle millions of rows.

Auto_TS: Auto_TimeSeries Automatically build multiple Time Series models using a Single Line of Code. Now updated with Dask. Auto_timeseries is a comp

AutoViz and Auto_ViML 519 Jan 3, 2023
MaD GUI is a basis for graphical annotation and computational analysis of time series data.
MaD GUI is a basis for graphical annotation and computational analysis of time series data.

MaD GUI Machine Learning and Data Analytics Graphical User Interface MaD GUI is a basis for graphical annotation and computational analysis of time se

Machine Learning and Data Analytics Lab FAU 10 Dec 19, 2022
Tool for producing high quality forecasts for time series data that has multiple seasonality with linear or non-linear growth.

Prophet: Automatic Forecasting Procedure Prophet is a procedure for forecasting time series data based on an additive model where non-linear trends ar

Facebook 15.4k Jan 7, 2023
Visualize classified time series data with interactive Sankey plots in Google Earth Engine
Visualize classified time series data with interactive Sankey plots in Google Earth Engine

sankee Visualize changes in classified time series data with interactive Sankey plots in Google Earth Engine Contents Description Installation Using P

Aaron Zuspan 76 Dec 15, 2022
PyPOTS - A Python Toolbox for Data Mining on Partially-Observed Time Series

A python toolbox/library for data mining on partially-observed time series, supporting tasks of forecasting/imputation/classification/clustering on incomplete multivariate time series with missing values.

Wenjie Du 179 Dec 31, 2022
Examples and code for the Practical Machine Learning workshop series

Practical Machine Learning Workshop Series Practical Machine Learning for Quantitative Finance Post conference workshop at the WBS Spring Conference D

CompatibL 21 Jun 25, 2022
Data science, Data manipulation and Machine learning package.
Data science, Data manipulation and Machine learning package.

duality Data science, Data manipulation and Machine learning package. Use permitted according to the terms of use and conditions set by the attached l

David Kundih 3 Oct 19, 2022
A python library for easy manipulation and forecasting of time series.
A python library for easy manipulation and forecasting of time series.

Time Series Made Easy in Python darts is a python library for easy manipulation and forecasting of time series. It contains a variety of models, from

Unit8 5.2k Jan 4, 2023
An open-source library of algorithms to analyse time series in GPU and CPU.
An open-source library of algorithms to analyse time series in GPU and CPU.

An open-source library of algorithms to analyse time series in GPU and CPU.

Shapelets 216 Dec 30, 2022
Comments
  • sklearn module import error

    sklearn module import error

    Describe the bug A clear and concise description of what the bug is. import error -> from time_series_transform.sklearn import * To Reproduce Steps to reproduce the behavior:

    1. Go to '...'
    2. Click on '....'
    3. Scroll down to '....'
    4. See error

    Expected behavior A clear and concise description of what you expected to happen.

    Screenshots If applicable, add screenshots to help explain your problem.

    Desktop (please complete the following information):

    • OS: [e.g. iOS]
    • Browser [e.g. chrome, safari]
    • Version [e.g. 22]

    Smartphone (please complete the following information):

    • Device: [e.g. iPhone6]
    • OS: [e.g. iOS8.1]
    • Browser [e.g. stock browser, safari]
    • Version [e.g. 22]

    Additional context Add any other context about the problem here.

    opened by allen-chiang 0
Releases(1.1.2)
Owner
Allen Chiang
Allen Chiang
GitHub Repository https://allen-chiang.github.io/Time-Series-Transformer/
Code for the TCAV ML interpretability project

Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV) Been Kim, Martin Wattenberg, Justin Gilmer, C

552 Dec 27, 2022
MICOM is a Python package for metabolic modeling of microbial communities

Welcome MICOM is a Python package for metabolic modeling of microbial communities currently developed in the Gibbons Lab at the Institute for Systems

57 Dec 21, 2022
Create large-scale ML-driven multiscale simulation ensembles to study the interactions

MuMMI RAS v0.1 Released: Nov 16, 2021 MuMMI RAS is the application component of the MuMMI framework developed to create large-scale ML-driven multisca

4 Feb 16, 2022
An easier way to build neural search on the cloud

Jina is geared towards building search systems for any kind of data, including text, images, audio, video and many more. With the modular design & multi-layer abstraction, you can leverage the effici

Jina AI 17k Jan 01, 2023
learn python in 100 days, a simple step could be follow from beginner to master of every aspect of python programming and project also include side project which you can use as demo project for your personal portfolio

learn python in 100 days, a simple step could be follow from beginner to master of every aspect of python programming and project also include side project which you can use as demo project for your

BDFD 6 Nov 05, 2022
Penguins species predictor app is used to classify penguins species created using python's scikit-learn, fastapi, numpy and joblib packages.

Penguins Classification App Penguins species predictor app is used to classify penguins species using their island, sex, bill length (mm), bill depth

Siva Prakash 3 Apr 05, 2022
A Python library for choreographing your machine learning research.

A Python library for choreographing your machine learning research.

AI2 270 Jan 06, 2023
scikit-learn models hyperparameters tuning and feature selection, using evolutionary algorithms.

Sklearn-genetic-opt scikit-learn models hyperparameters tuning and feature selection, using evolutionary algorithms. This is meant to be an alternativ

Rodrigo Arenas 180 Dec 20, 2022
Interactive Web App with Streamlit and Scikit-learn that applies different Classification algorithms to popular datasets

Interactive Web App with Streamlit and Scikit-learn that applies different Classification algorithms to popular datasets Datasets Used: Iris dataset,

Samrat Mitra 2 Nov 18, 2021
PLUR is a collection of source code datasets suitable for graph-based machine learning.

PLUR (Programming-Language Understanding and Repair) is a collection of source code datasets suitable for graph-based machine learning. We provide scripts for downloading, processing, and loading the

Google Research 76 Nov 25, 2022
ETNA is an easy-to-use time series forecasting framework.

ETNA is an easy-to-use time series forecasting framework. It includes built in toolkits for time series preprocessing, feature generation, a variety of predictive models with unified interface - from

Tinkoff.AI 674 Jan 07, 2023
Kubeflow is a machine learning (ML) toolkit that is dedicated to making deployments of ML workflows on Kubernetes simple, portable, and scalable.

SDK: Overview of the Kubeflow pipelines service Kubeflow is a machine learning (ML) toolkit that is dedicated to making deployments of ML workflows on

Kubeflow 3.1k Jan 06, 2023
My project contrasts K-Nearest Neighbors and Random Forrest Regressors on Real World data

kNN-vs-RFR My project contrasts K-Nearest Neighbors and Random Forrest Regressors on Real World data In many areas, rental bikes have been launched to

1 Oct 28, 2021
Data from "Datamodels: Predicting Predictions with Training Data"

Data from "Datamodels: Predicting Predictions with Training Data" Here we provid

Madry Lab 51 Dec 09, 2022
Time series changepoint detection

changepy Changepoint detection in time series in pure python Install pip install changepy Examples from changepy import pelt from cha

Rui Gil 92 Nov 08, 2022
Machine Learning toolbox for Humans

Reproducible Experiment Platform (REP) REP is ipython-based environment for conducting data-driven research in a consistent and reproducible way. Main

Yandex 663 Dec 31, 2022
Python Machine Learning Jupyter Notebooks (ML website)

Python Machine Learning Jupyter Notebooks (ML website) Dr. Tirthajyoti Sarkar, Fremont, California (Please feel free to connect on LinkedIn here) Also

Tirthajyoti Sarkar 2.6k Jan 03, 2023
ETNA – time series forecasting framework

ETNA Time Series Library Predict your time series the easiest way Homepage | Documentation | Tutorials | Contribution Guide | Release Notes ETNA is an

Tinkoff.AI 675 Jan 08, 2023
SynapseML - an open source library to simplify the creation of scalable machine learning pipelines

Synapse Machine Learning SynapseML (previously MMLSpark) is an open source library to simplify the creation of scalable machine learning pipelines. Sy

Microsoft 3.9k Dec 30, 2022
DistML is a Ray extension library to support large-scale distributed ML training on heterogeneous multi-node multi-GPU clusters

DistML is a Ray extension library to support large-scale distributed ML training on heterogeneous multi-node multi-GPU clusters

27 Aug 19, 2022