This document outlines a 4-phase process to analyze water quality data:
1) Identify trends and patterns in historical water quality indices.
2) Develop algorithms to identify seasonal patterns in pollution concentrations.
3) Use sliding windows and deep learning to detect anomalies in water quality data.
4) Determine the best time series models to model and forecast pollution levels, using GARCH models.
This document outlines a 4-phase process to analyze water quality data:
1) Identify trends and patterns in historical water quality indices.
2) Develop algorithms to identify seasonal patterns in pollution concentrations.
3) Use sliding windows and deep learning to detect anomalies in water quality data.
4) Determine the best time series models to model and forecast pollution levels, using GARCH models.
This document outlines a 4-phase process to analyze water quality data:
1) Identify trends and patterns in historical water quality indices.
2) Develop algorithms to identify seasonal patterns in pollution concentrations.
3) Use sliding windows and deep learning to detect anomalies in water quality data.
4) Determine the best time series models to model and forecast pollution levels, using GARCH models.
Identifying the persistency of water quality Information Gathering and indices and data summarization using Achieve Research Identification of persistency Markov chain model. Objective One (RO1) and trend of water quality Determining the trend of water quality indices indices at individual stations include i) testing the serial correlation effect ii) applying the MK test; iii) estimating the slope of a trend using Theil-Sen’s estimator
Phase 2 Computational algorithm will be developed in
order to estimate parameter of different Achieve Research Identification of seasonal harmonic of Fourier series. Objective Two (RO2) pattern in the water pollution The seasonal period of water quality will be concentration level produced
Phase 3 Univariate time series data will be
transformed into a shift approach called the Detection of anomalies in sliding or rolling window. water quality data using deep The sliding window approach to frame a Achieve Research learning approach dataset of the time series into independent Objective Three and dependent variables. (RO3) This mechanism allows all the time series data to be restructured to attribute value as to meet the requirement of the supervised (deep) learning approach.
Phase 4 Identify the stochastic processes implicit in
the time series, taking into account feedback Determining the best time with lag structure. series model in modelling GARCH model will be applied to capture such Achieve Research characteristics in the water quality time series Objective Four and forecasting water data (RO4) pollution centration level
Phase 4 The models will be assessed using a few criteria such
as mean squares error (MSE) and root means Model Validation squares error (RMSE).