Weather Forecasting
(Training Recurrent Neuro-Fuzzy System Using Two Novel Population-Based Algorithms for
Temperature Forecasting)
1. Introduction to the Proposed Summer Research
This proposal is intended to address research challenges in Weather Forecasting
specifically temperature forecasting a new interdisciplinary research area that
spans the Meteorology, Climatology, Atmospheric physics, Deep learning etc.
Of the atmospheric science and Computer Science.
Weather simply refer to the condition of the air on earth at a given place and
time. The application of science and technology to predict the state of the
atmosphere for a future time and a given location is so important due to its
effectiveness in hunam life.
A new structure of a Mamdani recurrent neuro-fuzzy system (MRNFS) model
will be used to temperature forecasting problem. The model considers two
recurrent properties, dynamic rules and the feedback connections which added
in the defuzzification layer. The operational parameters of this model are
trained using hybrid learning algorithm in which gradient descent (GD)
algorithm is used to train the output membership functions (MFs) values and
two novel population-based algorithms consist of the improved version of
honey bees optimization (HBO) and breeding swarms (BS) algorithm are used
to train the antecedent parameters of MRNFS model. The trained MRNFS is
then used to predict the future weather conditions.
2. Research Objectives and Approaches: Modeling and Simulations
Weather forecasts are made by collecting as much data as possible about the
current state of the atmosphere (particularly the temperature, humidity and
wind) and using understanding of atmospheric processes (through meteorology)
to determine how the atmosphere evolves in the future.
Predicting the weather is hard , but major weather events like earthquakes,
hurricanes, tornadoes, and typhoons cost billions of dollars each year in damage
.So we can try develop a better predicting technique which can save nationsand
lives. This is a real world problem which needs to be solved to save millions of
lives each year.
In this reasearch work i will be using Mamdani recurrent neuro-fuzzy system
(MRNFS).Evolutionary algorithms which inspired form biological science have
been applied to many complex optimization problems solving in recent years,
such as the optimal design of neural networks and fuzzy systems, also Fuzzy
inference system and neural network are other kinds of artificial intelligent
approaches which both are universal estimators. Neuro-fuzzy systems or neural
fuzzy systems (NFS) combine benefits of fuzzy inference and neural learning to
solve the problems of system modeling.
However, a major disadvantage of existing NFSs is that their application is
limited to static problems as a result of their internal feedforward network
structure. Recurrent neuro-fuzzy systems (RNFSs) incorporate feedback in their
topologies so they can overcome disadvantages of feedforward networks; also
system dynamics could be learned without much knowledge. The feedback loop
can be put in the premise part [1], in the consequence part [2], or in the final
output of fuzzy system [3]. If these feedback loops are added to membership
functions (MFs), it is said that this network is the NFS with dynamic fuzzy
rules.
3. Project Timeline
Week 1 Understanding the structure of Temperature Forecasting
System
Week 2 Implementing HoneyBee Optimization and testing on
sample data sets
Week 3 Learning Improved Hbo Algorithm
Week 4 Breeding Swarms Algorithm
Week 5 Learning Algorithm for Mamdani Recurrent Neuro
Fuzzy System Model
Week 6 Implementation of the model
4. Expected Outcome of the Project
A new effective technique to forecast the temperature using various algorithms.
This approach is able to determine the nonlinear relationship that exists between
the historical data (temperature, wind speed, humidity, etc.,) supplied to the
system during the training phase and on that basis, make a prediction of what
the temperature would be in future.
