Anamoly Detection for IOT Devices

Objective

Anomaly detection issue for time arrangement can be planned as discovering exception information guides relative toward some norm or common sign. Our center will be from a machine persopective, for example, surprising spikes, level move highlighting disintegrating soundness of a machine.

Dataset

Dataset contains 4 columns as follows:-

  • Datetime - 10 mins time interval of accelerometer data
  • 4-Bearings - Contains reading of devices

Anamoly Detection using Prophet Time Series Model Workflow

Prophet is a procedure for forecasting time series data based on an additive model where non-linear trends fit with yearly, weekly, daily, seasonality and holiday effects. It works best with time series that have strong seasonal effects and several seasons of historical data. Time-series anomaly detection is a feature used to identify unusual patterns that do not conform to expected behavior, called outliers.

Stock Forecasting

Data Preprocessing

  • Column Filter convert multivariate data into univariate for prophet model
Stock Forecasting
  • Output Univariate data
Stock Forecasting

Data Modeling

  • Prophet Model for anomaly detication using mean as threshold value

General Section of Prophet Model

  • Set Datetime column in DS column field
  • Y is the target variable. Set it to the reading of bearings
  • Set Growth as linear or logistic
  • We are using prophet model so that it is self-sufficient to select seasonality in auto mode
  • Set mode of seasonality as additive or multiplicative
  • Set confidence Interval (0 to 1) which gives a range of plausible values for the parameter of interest.
Stock Forecasting

Future Data section of Prophet model

  • FUTURE PERIOD block gives the number of steps we want to predict
Stock Forecasting
  • SQL set mean column to set threshold
Stock Forecasting

Model prediction

  • Threshold to compare anomaly
Stock Forecasting