Community Blog Project Showcase | Seattle Weather Forecasting Via Artificial Neural Network

Project Showcase | Seattle Weather Forecasting Via Artificial Neural Network

This project is from the team Sam&Ben, which was awarded with the Innovation Award in the Global AI Innovation Challenge 2021 - Intelligent Weather Forecast for Better life.

This project is from the team Sam&Ben, which was awarded with the Innovation Award in the Global AI Innovation Challenge 2021 - Intelligent Weather Forecast for Better life.

Project Introduction

Weather forecasting has become one of the intelligent technologies to predict temperature, humidity, rainfall and wind speed. It has been widely used across the globe. Weather prediction plays a vital role in every sector of our lives, including business, tourism, agriculture, power, airport, and others (Jaseena & Kovoor, 2020). Besides, weather prediction has changed our lives; it also keeps us safe from the warning of disasters such as floods, storms, and tornadoes (Jaseena & Kovoor, 2020). For instance, in agriculture industries, weather prediction has become essential to farmers, to produce all crops and livestock depending on the prior information about the weather. Hence, weather forecasting and data analysis require a group of accurate quantitative data. With the improvement and development of technology in recent times, weather forecasting has been improving and predicting the future climate more accurately. These technologies include the Internet of Things, Cloud Computing (Jaseena & Kovoor, 2020).

Hence, this project proposes using an Artificial Neural Network as a prediction model for weather forecasting.


Solution and Project Value

Seattle is one of the rainiest cities in the United States. The city has approximately 152 rainy days in a year, which means it has more days of rain compared to other states in the country. According to the Seattle Office of Emergency Management, the heavy rains have been a disastrous threat to the urban area. The residential flooding has greatly affected the agricultural and tourism industry, it also worsens the economy. Therefore, weather forecasting plays a crucial role to prevent disastrous threats. Hence, this project proposes a prediction model to predict the weather condition via Artificial Neural Networks. Weather forecasting aims in applying technologies from a wide array of scopes to narrow down as to when it would rain and when there exists a high/low temperature. Seattle is most famous for how often it rains (NOAA). A dataset collected at the Seattle-Tacoma International Airport contains five columns: Date, Prep, Tmax, Tmin, and Rain. This dataset compiled by NOAA was used to successfully welcome the objectives of this hand of work. The prediction model was made by means of deploying MATLAB to the rescue. Artificial Neural Network, together with Levenberg-Marquardt Algorithm were of good use here

1 Peter 4: 10: '“Each of you should use whatever gift you have received to serve others, as faithful stewards of God’s grace in its various forms”. We aim to procure a caliber of lucidity among individuals when it comes to all things about the weather, concretely that of Seattle, and to inspire and bring substantial value to the people.

Our solution to develop a weather forecasting prediction model using Artificial Neural Networks is:

  1. Build a model to predict on which exact day it would rain in Seattle based on information/data from 1948 to 2017.
  2. Strike a correlation between the minimum and maximum temperature thresholds in Seattle.
  3. Create awareness of how Machine Learning aids in weather forecasting.

Technology Highlights

The Levenberg-Marquart algorithm comprises the Gradient Descent Method and the Gauss-Newton Method. The Gradient Descent Method involves updating the parameter values in the “downhill” direction, i.e., in the direction opposite to the gradient of the objective function while the Gauss-Newton Method involves the minimization of the sum of squares objective function. The size of problems is usually curtailed to each method – for problems with thousands of parameters, Gradient descent methods have an upper hand, for problems that are moderately sized the Gauss-Method is preferred. The regression of the model is 0.9116 and the best training performance is 1888.9345 at epoch 1000. The range of the hidden neuron number is from 1 to 20, and the best performance of the training process with the most accurate results is at 1st hidden neuron. The regression of the sample of training, validation, and testing results in 0.86596, 0.87327, and 0.86533 respectively. The overall regression is 0.86699. Root Mean Square Error from the comparison error between the predicted results and actual results in order to achieve the most accurate result. The graph illustrates the differences between the actual results and predicted results, whereas the ‘yellow line’ represents actual results and the ‘blue’ line as predicted results.

Alibaba Cloud Products Used

About the Developer

We are two students hailing from the hinterlands of Malaysia:

  1. Samuel Otabir, is an upcoming electrical/electronic engineer from UCSI University. Samuel is very passionate about all things data. Samuel has worked on various projects, ranging from building smart machine models, websites, and applications. Samuel plans to give back to the world by any means possible to enhance the livelihood of people.
  2. Benjamin Wong Wei Herng, is an upcoming mechatronics engineer from UCSI University. Benjamin has a vigorous affinity for all learning cognate to Robotics systems, Artificial Intelligence, and IoT futures. Benjamin has withal accomplished tasks like building an autonomous robotic waiter in a hospital and developing models pertaining to head movements in autonomous cars by the deployment of ANN with different training algorithms.
0 0 0
Share on

Alibaba Cloud Project Hub

113 posts | 22 followers

You may also like