How to use XGBoost and InluxDB for time series forecasting

XGBoost is a popular open source machine learning library that can be used to solve a variety of prediction problems. One needs to understand how to use it with InfluxDB for time series forecasting.

Translator | Li Rui

Reviewer | Sun Shujuan

XGBoost is an open source machine learning library that implements an optimized distributed gradient boosting algorithm. XGBoost uses parallel processing for fast performance, handles missing values ​​well, performs well on small datasets, and prevents overfitting. All these advantages make XGBoost a popular solution for regression problems such as prediction.

Forecasting is mission-critical for various business goals such as predictive analytics, predictive maintenance, product planning, budgeting, etc. Many forecasting or forecasting problems involve time series data. This makes XGBoost an excellent partner for the open source time series database InfluxDB.

This tutorial will learn how to use XGBoost’s Python package to predict data from the InfluxDB time series database. You will also use the InfluxDB Python client library to query data from InfluxDB and convert the data to a Pandas DataFrame to make it easier to work with time series data before making predictions. Additionally, the advantages of XGBoost will be discussed in more detail.

1. Requirements

This tutorial is performed on a macOS system with Python 3 installed through Homebrew. It is recommended to set up additional tools such as virtualenv, pyenv or conda-env to simplify Python and client installation. Otherwise, its full requirements are as follows:

• influxdb-client=1.30.0
• pandas = 1.4.3
• xgboost>=1.7.3
• influxdb-client>=1.30.0
• pandas>=1.4.3
• matplotlib>=3.5.2
• sklearn>=1.1.1

This tutorial also assumes that you have a free tier InfluxDB cloud account and that you have created a bucket and a token. Think of a bucket as the database or the highest level of data organization in InfluxDB. In this tutorial, a bucket named NOAA will be created.

2. Decision trees, random forests and gradient enhancement

In order to understand what XGBoost is, you must understand decision trees, random forests and gradient enhancement. Decision trees are a supervised learning method consisting of a series of feature tests. Each node is a test, and all nodes are organized in a flowchart structure. Branches represent conditions that ultimately determine which leaf label or class label is assigned to the input data.

# Decision trees in machine learning are used to determine whether it will rain tomorrow. Edited to show the components of a decision tree: leaves, branches, and nodes.

The guiding principle behind decision trees, random forests, and gradient boosting is that multiple "weak learners" or classifiers work together to make strong predictions.

Random forest contains multiple decision trees. Each node in a decision tree is considered a weak learner, and each decision tree in a random forest is considered one of many weak learners in a random forest model. Typically, all data is randomly divided into subsets and passed through different decision trees.

Gradient boosting using decision trees and random forests is similar, but the way they are structured is different. Gradient boosted trees also contain decision tree forests, but these decision trees are additionally constructed and all the data is passed through the decision tree ensemble. Gradient boosting trees may consist of a set of classification trees or regression trees, with classification trees for discrete values ​​(such as cats or dogs). Regression trees are used for continuous values ​​(e.g. 0 to 100).

3. What is XGBoost?

Gradient boosting is a machine learning algorithm used for classification and prediction. XGBoost is just an extreme type of gradient boosting. At its extreme, gradient boosting can be performed more efficiently through the power of parallel processing. The image below from the XGBoost documentation illustrates how gradient boosting can be used to predict whether someone will like a video game.

#Two decision trees are used to decide whether someone is likely to like a video game. Add the leaf scores from both trees to determine which person is most likely to enjoy the video game.

Some advantages of XGBoost:

• Relatively easy to understand.
• Suitable for small, structured and regular data with few features.

Some disadvantages of XGBoost:

• Easy to overfitting and sensitive to outliers. It might be a good idea to use materialized views of time series data in XGBoost for forecasting.
• Performs poorly on sparse or unsupervised data.

4. Use XGBoost for time series prediction

The air sensor sample data set used here is provided by InfluxDB. This dataset contains temperature data from multiple sensors. A temperature forecast is being created for a single sensor with data like this:

Use the following Flux code to import the dataset and filter for a single time series. (Flux is the query language of InfluxDB)

import "join"

import "influxdata/influxdb/sample"

//dataset is regular time series at 10 second intervals

data = sample.data(set: "airSensor")

|> filter(fn: (r) => r._field == "temperature" and r.sensor_id == "TLM0100")

Random forest and gradient boosting can be used for time series forecasting, but they require converting the data to supervised learning. This means that the data must be moved forward in a sliding window approach or a slow-moving approach to convert the time series data into a supervised learning set. The data can also be prepared with Flux. Ideally, some autocorrelation analysis should be performed first to determine the best method to use. For the sake of brevity, the following Flux code will be used to move data at a regular interval.

import "join"

import "influxdata/influxdb/sample"

data = sample.data(set: "airSensor")

|> ; filter(fn: (r) => r._field == "temperature" and r.sensor_id == "TLM0100")

shiftedData = data

|> timeShift(duration : 10s, columns: ["_time"] )

join.time(left: data, right: shiftedData, as: (l, r) => ({l with data: l._value, shiftedData : r._value}))

|> drop(columns: ["_measurement", "_time", "_value", "sensor_id", "_field"])

Swipe left and right to see the complete code

If you want to add additional lag data to the model input, you can follow the following Flux logic instead.

import "experimental"

import "influxdata/influxdb/sample"

data = sample.data(set: "airSensor")

|> ; filter(fn: (r) => r._field == "temperature" and r.sensor_id == "TLM0100")



shiftedData1 = data

|> timeShift(duration: 10s, columns: ["_time"] )

|> set(key: "shift" , value: "1" )



shiftedData2 = data

|> timeShift(duration: 20s , columns: ["_time"] )

|> set(key: "shift" , value: "2" )



shiftedData3 = data

|> timeShift(duration: 30s , columns: ["_time"] )

|> set( key: "shift" , value: "3")



shiftedData4 = data

|> timeShift(duration: 40s , columns: ["_time"] )

|> set(key: "shift" , value: "4")


##union(tables: [shiftedData1, shiftedData2, shiftedData3, shiftedData4])

|> pivot(rowKey:["_time"], columnKey: ["shift"], valueColumn: "_value")

|> drop(columns: ["_measurement", "_time", "_value", "sensor_id", "_field"])

// remove the NaN values

|> limit(n:360)

|> tail(n: 356)

In addition, we must use forward validation to train the algorithm. This involves splitting the data set into a test set and a training set. Then use XGB Regressor to train the XGBoost model and use the fitting method to predict. Finally, we use the mean absolute error (MAE) to determine the accuracy of the forecast. For a delay of 10 seconds, the calculated mean absolute error (MAE) is 0.035. We can interpret this to mean that 96.5% of the predictions are very correct. The figure below shows the comparison between the predictions we get from XGBoost and the expected values ​​from the train/test split.

The following is the complete script. This code is mostly taken from the tutorial here.

import pandas as pd

from numpy import asarray

from sklearn.metrics import mean_absolute_error

from xgboost import XGBRegressor

from matplotlib import pyplot

from influxdb_client import InfluxDBClient

from influxdb_client.client.write_api import SYNCHRONOUS



# query data with the Python InfluxDB Client Library and transform data into a supervised learning problem with Flux

client = InfluxDBClient(url="https://us-west-2-1.aws.cloud2.influxdata.com", token="NyP-HzFGkObUBI4Wwg6Rbd-_SdrTMtZzbFK921VkMQWp3bv_e9BhpBi6fCBr_0-6i0ev32_XWZcmkDPsearTWA==", org="0437f6d51b579000")



# write_api = client.write_api(write_optinotallow=SYNCHRONOUS)

query_api = client.query_api()

df = query_api.query_data_frame('import "join"'

'import "influxdata/influxdb/sample"'

'data = sample.data(set: "airSensor")'

'|> filter(fn: (r) => r._field == "temperature" and r.sensor_id == "TLM0100")'

'shiftedData = data'

'|> timeShift(duration: 10s , columns: ["_time"] )'

'join.time(left: data, right: shiftedData, as: (l, r) => ({l with data: l._value, shiftedData: r._value}))'

'|> drop(columns: ["_measurement", "_time", "_value", "sensor_id", "_field"])'

'|> yield(name: "converted to supervised learning dataset")'

)

df = df.drop(columns=['table', 'result'])

data = df.to_numpy()



# split a univariate dataset into train/test sets

def train_test_split(data, n_test):

return data[:-n_test:], data[-n_test:]



# fit an xgboost model and make a one step prediction

def xgboost_forecast(train, testX):

# transform list into array

train = asarray(train)

# split into input and output columns

trainX, trainy = train[:, :-1], train[:, -1]

# fit model

model = XGBRegressor(objective='reg:squarederror', n_estimators=1000)

model.fit(trainX, trainy)

# make a one-step prediction

yhat = model.predict(asarray([testX]))

return yhat[0]



# walk-forward validation for univariate data

def walk_forward_validation(data, n_test):

predictions = list()

# split dataset

train, test = train_test_split(data, n_test)

history = [x for x in train]

# step over each time-step in the test set

for i in range(len(test)):

# split test row into input and output columns

testX, testy = test[i, :-1], test[i, -1]

# fit model on history and make a prediction

yhat = xgboost_forecast(history, testX)

# store forecast in list of predictions

predictions.append(yhat)

# add actual observation to history for the next loop

history.append(test[i])

# summarize progress

print('>expected=%.1f, predicted=%.1f' % (testy, yhat))

# estimate prediction error

error = mean_absolute_error(test[:, -1], predictions)

return error, test[:, -1], predictions



# evaluate

mae, y, yhat = walk_forward_validation(data, 100)

print('MAE: %.3f' % mae)



# plot expected vs predicted

pyplot.plot(y, label='Expected')

pyplot.plot(yhat, label='Predicted')

pyplot.legend()

pyplot.show()

五、结论

希望这篇博文能够激励人们利用XGBoost和InfluxDB进行预测。为此建议查看相关的报告，其中包括如何使用本文描述的许多算法和InfluxDB来进行预测和执行异常检测的示例。

原文链接：https://www.infoworld.com/article/3682070/time-series-forecasting-with-xgboost-and-influxdb.html

The above is the detailed content of How to use XGBoost and InluxDB for time series forecasting. For more information, please follow other related articles on the PHP Chinese website!