Answer To: MS Seminar I: Rules, results and some other topics Micha l Ma lafiejski 19/01/2021 General grading...
Swapnil answered on Jan 21 2021
Homework 7 8 9 10/Homework 10/Homework 10.pdf
Homework X: Draft Thesis Statement
My thesis is as follows. It will basically contain the following points.
What do you claim:
A Multilayer Perceptron with Back-Propagation algorithm, a Dynamic Neural Network and
two hybrid ANNs. It claimed that provided better results than the MLP-BP model. These
hybrid neural network models associate the Fourier series with the ANN. This association
consists of modelling by ANN the residuals of the Fourier series forecast. As an example,
they claimed that simplistic Gaussian residual assumptions were not appropriate for demand
forecasting. The support vector regression, as claimed as one of the best methods for short-
term water predictions. Additionally, they built a Fourier time series process that allowed
enhancing the prediction. The combination of both processes was used to build a tool that
provides several contributions, namely elimination of errors and the part of bias inherent in
a fixed regression structure whenever new time series data are incorporated into the analysis.
How do you want to solve the problem:
The water demand forecasting problem in order to evaluate the predictive uncertainty in
water networks. Based on the application of the MCP, they were able to combine more than
one forecasting model and to obtain a probability distribution of short-term demands, which
depended on the values forecasted by each model combined. That distribution was then
used to infer the future demand and predictive uncertainty. The water demand forecasting
problem in order to remove the need for determining annual seasonality and extending the
prediction. The method we present deals with the problem of forecasting with horizon and
at the same time, forecasting the next 1440 values of the series.
Which techniques / technologies / methods do you plan
to use to solve the problem:
A wide variety of technologies has been deployed that have the potential to change the
paradigm of the management of water distribution networks, turning them into smart water
networks. The smart water networks consist of a large number of devices that measure a
wide range of parameters present in distribution networks in an automatic and continuous
way. We have used the Machine learning technology with different types of algorithms. That
will solve the issues of the water management.
Write precise goals and expected outcomes:
The goal of this section is to present the different locations that were used in the study as well
as the method used to pre-process the data before performing our analysis. Moreover, the
1
section includes the description of several concepts related to the study, such as the trends
of seasonalities, the input/output patterns, and the proposed algorithm. The following
outcomes will come for the Water Consumption Prediction:
• To integrate and demonstrate the Water Consumption Prediction.
• To demonstrate 4 integrated solutions.
• To establish and guard integration and standardization aspects for the Water Consumption
Prediction.
• To establish the Water Consumption Prediction business cases, deployment potential and
market uptake routes.
2
Homework 7 8 9 10/Homework 10/main.tex\documentclass[12pt]{article}
% Formatting
\usepackage[utf8]{inputenc}
\usepackage[margin=1in]{geometry}
\usepackage[titletoc,title]{appendix}
\setlength{\parindent}{4em}
\begin{document}
\section*{Homework X: Draft Thesis Statement}
My thesis is as follows. It will basically contain the following points.
\section*{What do you claim:}
A Multilayer Perceptron with Back-Propagation algorithm, a Dynamic Neural Network and two hybrid ANNs. It claimed that provided better results than the MLP-BP model. These hybrid neural network models associate the Fourier series with the ANN. This association consists of modelling by ANN the residuals of the Fourier series forecast. As an example, they claimed that simplistic Gaussian residual assumptions were not appropriate for demand forecasting. The support vector regression, as claimed as one of the best methods for short-term water predictions. Additionally, they built a Fourier time series process that allowed enhancing the prediction. The combination of both processes was used to build a tool that provides several contributions, namely elimination of errors and the part of bias inherent in a fixed regression structure whenever new time series data are incorporated into the analysis.
\section*{How do you want to solve the problem:}
The water demand forecasting problem in order to evaluate the predictive uncertainty in water networks. Based on the application of the MCP, they were able to combine more than one forecasting model and to obtain a probability distribution of short-term demands, which depended on the values forecasted by each model combined. That distribution was then used to infer the future demand and predictive uncertainty. The water demand forecasting problem in order to remove the need for determining annual seasonality and extending the prediction. The method we present deals with the problem of forecasting with horizon and at the same time, forecasting the next 1440 values of the series.
\section*{Which techniques / technologies / methods do you plan to use to solve the problem:}
A wide variety of technologies has been deployed that have the potential to change the paradigm of the management of water distribution networks, turning them into smart water networks. The smart water networks consist of a large number of devices that measure a wide range of parameters present in distribution networks in an automatic and continuous way. We have used the Machine learning technology with different types of algorithms. That will solve the issues of the water management.
\section*{Write precise goals and expected outcomes:}
The goal of this section is to present the different locations that were used in the study as well as the method used to pre-process the data before performing our analysis. Moreover, the section includes the description of several concepts related to the study, such as the trends of seasonalities, the input/output patterns, and the proposed algorithm.
The following outcomes will come for the Water Consumption Prediction:
\item • To integrate and demonstrate the Water Consumption Prediction.
\item • To demonstrate 4 integrated solutions.
\item • To establish and guard integration and standardization aspects for the Water Consumption Prediction.
\item • To establish the Water Consumption Prediction business cases, deployment potential and market uptake routes.
\end{document}
Homework 7 8 9 10/Homework 7/Homework 7.pdf
Homework VII: Motivation and Idea
Research Topic: A Short-Term Data Based Water Con-
sumption Prediction Approach
Motivation:
A smart water network consists of a large number of devices that measure a wide range of
parameters present in distribution networks in an automatic and continuous way. Among
these data, you can find the flow, pressure, or tantalizer measurements that, when processed
with appropriate algorithms, allow for leakage detection at an early stage. These algorithms
are mainly based on water demand forecasting. Different approaches for the prediction of
water demand are available in the literature. Although they present successful results at
different levels, they have two main drawbacks: the inclusion of several seasonality is quite
cumbersome, and the fitting horizons are not very large. With the aim of solving these
problems, we present the application of pattern similarity-based techniques to the water
demand forecasting problem. The use of these techniques removes the need to determine the
annual seasonality and, at the same time, extends the horizon of prediction to 24 h. The
algorithm has been tested in the context of a real project for the detection and location of
leaks at an early stage by means of demand forecasting, and good results were obtained,
which are also presented in this paper.
Recently, a wide variety of technologies has been deployed that have the potential to
change the paradigm of the management of water distribution networks, turning them into
smart water networks (SWN). An SWN consists of a large number of devices that measure
a wide range of parameters present in distribution networks in an automatic and continuous
way. Among these data, you can find flow, pressure, or tantalizer measurements that, when
processed with the appropriate algorithms, allow for the detection of leakages at an early
stage. These algorithms are mainly based on water demand forecasting.
The approach in dealt with the improvement of hourly stream flow forecasting perfor-
mance by proposing a methodology for the development of ANN-based models where an en-
hanced learning strategy was used. For this purpose, the authors considered several ANNs:
radial basis function network, self-organizing map, SVM, and back propagation network.
Used real data acquired from the city of Milan to test an approach for short-term prediction
of water consumption. The approach was based on a two-stage learning schema. While
the first stage concerned time-series data clustering, the second stage related to SVM for
regression.
Some example scenarios:
As an example, they claimed that simplistic Gaussian residual assumptions were not ap-
propriate for demand forecasting. For example, the Water Demand Forecast introduced in
uses a moving window of three weeks to obtain average parameters for similar days giving
a forecast horizon of 24 h, regardless of the data sample frequency. Another approach given
1
in used a generalized regression neural network, which makes use of Gaussian radial basis
functions to approximate a target function in multidimensional spaces.
A high frequency random component that overlapped the signal. Due to this random
fluctuation, the FOB value was increased to 0.51. However, thanks to the pattern followed
by the forecasted and measured values, which were similar, the original signal was well
predicted. This is the reason why RMSE remained small (RMSE = 2.8 L/min) even though
FOB was high
The approach in dealt with the improvement of hourly streamflow forecasting perfor-
mance by proposing a methodology for the development of ANN-based models where an en-
hanced learning strategy was used. For this purpose, the authors considered several ANNs:
radial basis function network, self-organizing map, SVM, and back propagation network.
Used real data acquired from the city of Milan to test an approach for short-term prediction
of water consumption. The approach was based on a two-stage learning schema. While
the first stage concerned time-series data clustering, the second stage related to SVM for
regression.
Idea of a problem:
We find different techniques to cope with...