AU2020103741A4 - Feed forward neural networks combined with extreme learning machine approach for large weather data - Google Patents
Feed forward neural networks combined with extreme learning machine approach for large weather data Download PDFInfo
- Publication number
- AU2020103741A4 AU2020103741A4 AU2020103741A AU2020103741A AU2020103741A4 AU 2020103741 A4 AU2020103741 A4 AU 2020103741A4 AU 2020103741 A AU2020103741 A AU 2020103741A AU 2020103741 A AU2020103741 A AU 2020103741A AU 2020103741 A4 AU2020103741 A4 AU 2020103741A4
- Authority
- AU
- Australia
- Prior art keywords
- data
- weather
- neural network
- feed forward
- network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/02—Instruments for indicating weather conditions by measuring two or more variables, e.g. humidity, pressure, temperature, cloud cover or wind speed
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Abstract
FEED FORWARD NEURAL NETWORKS COMBINED WITH EXTREME LEARNING
MACHINE APPROACH FOR LARGE WEATHER DATA
ABSTRACT
Weather forecasting is the application of recent technology and science to find out the state of
atmosphere for the purpose of determining the future. The precise weather forecasting is
significant in today's world because the sectors like agriculture and industries will depend on it.
As the weather condition is dynamic and non-linear process, the proposed method addresses the
artificial neural network to classify and predict the weather conditions and it is one of the best
methods comparing to conventional methods. The neural network models will help to support the
various training or learning algorithm. To train the neural network, the back propagation is one of
the important algorithms for weather forecasting.
1
Description
Description
Field of Invention:
This field of invention addresses the feed forward neural networks combined with the extreme machine learning approach for the large volume of weather data. The proposed method presents the artificial neural network for the purpose of weather predictions. The neural network model supports the various training or learning algorithms.
Background of invention:
The weather parameters like highest and lowest temperature, humidity and the average wind speed of the rice research station is used. The ensemble model is better comparing to other neural network models such as multi-layered perceptron network, Elman recurrent neural network, radial basis function network, Hopfield model and the regression techniques which is stated by the Imran Maqsood et al. (2004). The model is helpful for performing the multi-class classification problems without any increase of the complexity. The satellite-based system is ignored for forecasting the weather conditions because this system is expensive and it requires the support system. The artificial neural network is used along with the time series features by Paras, Sanjay Mathur, Avinash Kumar, and Mahesh Chandra et al. Nekoukar et al., 2010 described that, the neural network is applicable to most of the aspects of prediction and radial based function neural networks is used efficiently for most of thefinancial time-series forecasting. The artificial neural network with the help of back propagation technique is studied by Ch. Jyosthna Devi, B. Syam Prasad Reddy, K. Vagdhan Kumar, B. Musala Reddy, N. Raja Nayak et al. (2012). The method is more efficient comparing to numerical differentiation. The Back propagation neural network is best and efficient method to know about forecast monsoon rainfall and also about the other weather parameter predictions over the small geographical region which is studied by the Gyanesh Shrivastava, Sanjeev Karmakar, Manoj Kumar, Pulak Guhathakurta et al. The back propagation neural network is given as the appropriate solution for the purpose offinding out the longer weather forecasting. The Levenberg Marquardt back propagation is the fastest algorithm among various back propagation algorithms which is stated by the Arti R. Naik, S.K. pathan et al. (2012). The number of models designed by using the numerical weather prediction in the present in the past. The fuzzy recurrent neural network on the basis of time series will be forecasted in order to solve the forecasting problems which is proposed by the Aliev et al., 2008. The meteorological method for forecasting is compared with the artificial neural network by KOSCAK et al., 2009 and also, he founded the performance of the artificial neural network with the high accuracy. The radial based neural network function which is used by the Nekoukar et al., 2010, for the purpose of forecasting the finance time-series and their result also shows the effectiveness and its feasibility. The prediction of rainfall in Chennai is done by Geetha and Selvaraj, 2011, and they have used the back propagation neural network models and as per their research, the monthly rainfall will be known with the help of artificial neural network model. The model can also be well performed by the training period.
Objects of the invention:
• The objective of the proposed method is about the weather forecasting with the help of Feed-Forward Artificial Neural Network method. • The other objective is that, data can be trained by using Levenberg-Marquardt algorithm, as it is fastest method comparing to other weather forecasting methods. • The application of artificial neural network is used for the purpose of weather classification
and prediction, trains the network with the help of back propagation algorithm.
Summary of the invention:
The weather status is the state of atmosphere at the specified time in the form of weather variables such as temperature, pressure, wind direction, etc. The weather warnings are significant in order to protect the life and property. The Forecasts on the basis of temperature and precipitations are essential to the field of agriculture and industry. The temperature forecasts are done with the help of collection of data related to the recent state of the atmosphere. The benefit of back propagation neural network method is that it can approximate the functions. The abrupt change in climatic conditions and alertness of natural disasters can be predicted by knowing the weather conditions. The neural network process will be similar to the process of human brain because it will access the knowledge via the learning and its knowledge will be stored within the interneuron connection.
The Levenberg-Marquardt algorithm is good and faster, as it will achieve better performance comparing to other algorithms in training. If the memory availability is enough in the small and medium sized network, then the LM training can be used.
The feed forward neural network will have either the layers or subgroups of the processing elements. Firstly, the training will be given in order to create the network object. Feed-forward network with a greater number of layers will learn some complex relationship in faster way. This neural network does not have any feedback and it allows the signal to travel in only one way. The feed forward artificial neural network is the straight forward method and that use the inputs and outputs. The input and output values are to be provided in order to initiate the values of weight and bias for the purpose of creating network and also to determine the size of the output layer. The multiple layers are present in the network and it consists of three layers. The first layer is the input layer, second layer is considered as hidden layer, and third layer is output layer.
The feed-forward topology, supervised learning and the back-propagation learning algorithm is used by the back propagation neural network model. It has good capability in application and also it will have its own limitation. It is the general-purpose learning algorithm and also it is powerful and has very good potential. This method is expensive because of the computational requirements for training. The two phases are present in the learning cycle. One of the cycles is to propagate the patterns of input via the network and other cycle is to adapt the output by changing the weights in the network. This is supervised learning method and for many of the inputs, the dataset of the desired output is required for the training set. This will be helpful for the feed-forward neural networks. The back-propagation method is the method of gradient-descent and this method will be helpful in adjusting the weights as per the error function. The weather prediction by using the back propagation neural network collects the data like temperature, pressure, wind, humidity, and direction. The weather forecasting is done by collection of previous and current data of the atmosphere. Later, this data will be used to train the neural network. The training process consist of four steps. 1. The training data is to be assembled. 2. The network object is to be created. 3. The network to be trained. 4. The network response is to be simulated to the new inputs.
Detailed Description of the Invention:
The weather forecasting is the complex task as it deals with the large volume of data which is collected from the different weather satellites from every day. The collection of data values, and finding out the various patterns in the observation table, and later working with the result for the exact weather prediction output. In order to prevent the disaster such as flood, etc., it is significant to know about the weather data which is collected and analyzed in the real-time. The future weather condition is estimated by predicting the weather and it is the state of atmosphere at the specified time in the form of weather parameters like temperature, wind, direction and pressure etc. The weather is the non-linear and dynamic process, as it varies for every day and also for every minute. The ANN is matured enough to the extent for the past few years. The neural network method will solve the problems of the non-linear neurons. The neuron which is connected to every other neuron through the link. So, this network is trained with the help of back-propagation algorithm and that follows the gradient descent method. The new technique will be proposed for the weather classification and forecasting by using the Feed Forward Neural Network. In this network model, the number of high non-linear neurons which is interconnected to form the network. The three layers are formed in the network such as input, hidden, and output layer. The neurons are connected by the links and it also consists of the weight. The weights are considered as the strong bonding for connections and that will exist between the neurons in the network. The artificial neural network is the system that will receive the input, process the data and then later gives the output with respect to the input. The input, hidden layer, and output layer is present in the multi-layer neural network. As the weather condition is the data-intensive process and the weather data set will be non-linear, so that, the prediction will be done in precise way by using the artificial neural network and that will be designed over the feed-forward neural network, as it is used widely for the purpose of forecasting the weather and financial. The benefit over the artificial neural network is to extract the data, trending the data, and also, they can predict the patterns which will not provide while training the data. The ANN is processed in three phases. 1. Training, 2. Validating, 3. Testing. The network will be trained with the help of input dataset. If the greater number of input data set is used for the purpose of training, the result will be in precise way and also it will have the greater number of data which is available for the training. In the back-propagation algorithm, the error can also propagate in the back way, and the weights can also adjust in order to reduce the error. In initial stage of the training process, the connection present between the neurons will be set to the weight values in random way. In the training process, the data of the input and output from the training data subset will be sent to the network. The difference present between the output values of training and actual data is to be calculated.
The Figure 1 describes about the weather forecasting using the Artificial Neural Network. The weather classification and prediction are done by using the back propagation neural network, as it targets to collect the data like weather parameters such as temperature, wind, direction, pressure, and humidity. The input will be fed to the back propagation neural network and also the data will be collected related to previous as well as recent state of the atmosphere. The data is later then use to train the neural network. The data will be fed to the input layer of the network and later all the data will be collected using the software as well as the sensors such as thermo hygro sensor and anemometer sensor. The pre-processing of data will be after done after the stage of data collection. The noise which is present in the data will be removed to get the better result of prediction. The feed forward neural network will be trained with the help of Levenberg Marquardt back propagation neural network for the purpose of weather prediction. The classification is later done with the help of neural network classification. The data is stored by using the weather sensor and then, the data will be fed to the input to the network. The data is then pre-processed for the purpose of removing the noise and it will be trained and validated. Later, the output is then classified such as weather conditions like sunny, cloudy, and rainy by using the decision tables. The proposed method for weather forecasting is done by using the feed forward artificial neural network and later this data will be trained by using back propagation algorithm to predict the future purpose and classify the data.
Claims (8)
1. The Accurate weather prediction and classification can be done using the feed forward artificial neural network because the weather data is non-linear and training can be given to the network using the back-propagation algorithm.
2. The new technique is proposed for the classification of weather and forecasting by using the LM back propagation feed forward neural network.
3. The method claim2, the data collection is to be done based on past and recent state of the atmosphere and these data is used to train the neural network.
4. The method claim2, the pre-processing of data is done to remove the noise for the purpose of getting better prediction result.
5. The method claims, the method used here is feed forward neural network and this network is trained by using the LM back propagation neural network to predict the future weather condition and later to classify the data.
6. The method claim2, the neural network classification is done in order to convey about the weather condition that whether the climate is rainy, sunny, and partly cloudy, etc.
7. The method claim, the data can also be saved using the weather sensors and this data can be later fed as the input to the network.
8. The method claim2, the data is trained and validated, after the stage of pre-processing is done to the data.
FEED FORWARD NEURAL NETWORKS COMBINED WITH EXTREME LEARNING 27 Nov 2020
MACHINE APPROACH FOR LARGE WEATHER DATA
Drawings: ` 2020103741
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
AU2020103741A AU2020103741A4 (en) | 2020-11-27 | 2020-11-27 | Feed forward neural networks combined with extreme learning machine approach for large weather data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
AU2020103741A AU2020103741A4 (en) | 2020-11-27 | 2020-11-27 | Feed forward neural networks combined with extreme learning machine approach for large weather data |
Publications (1)
Publication Number | Publication Date |
---|---|
AU2020103741A4 true AU2020103741A4 (en) | 2021-02-11 |
Family
ID=74502381
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
AU2020103741A Ceased AU2020103741A4 (en) | 2020-11-27 | 2020-11-27 | Feed forward neural networks combined with extreme learning machine approach for large weather data |
Country Status (1)
Country | Link |
---|---|
AU (1) | AU2020103741A4 (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112906871A (en) * | 2021-03-24 | 2021-06-04 | 临沂大学 | Temperature prediction method and system based on hybrid multilayer neural network model |
CN113011092A (en) * | 2021-03-15 | 2021-06-22 | 广东电网有限责任公司清远供电局 | Meteorological environment monitoring method, system, electronic equipment and storage medium |
CN113192034A (en) * | 2021-04-30 | 2021-07-30 | 西安理工大学 | Mixed dye solution concentration detection method based on back propagation neural network |
CN113435628A (en) * | 2021-05-28 | 2021-09-24 | 淮阴工学院 | Medium-and-long-term runoff prediction method and system based on linear discriminant analysis and IALO-ELM |
CN113494527A (en) * | 2021-07-30 | 2021-10-12 | 哈尔滨工业大学 | Constant force control method based on electromagnetic auxiliary type constant force spring support |
CN114994794A (en) * | 2022-06-24 | 2022-09-02 | 昆明学院 | Cloud particle phase state growth method for cloud cluster non-detection data area |
-
2020
- 2020-11-27 AU AU2020103741A patent/AU2020103741A4/en not_active Ceased
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113011092A (en) * | 2021-03-15 | 2021-06-22 | 广东电网有限责任公司清远供电局 | Meteorological environment monitoring method, system, electronic equipment and storage medium |
CN112906871A (en) * | 2021-03-24 | 2021-06-04 | 临沂大学 | Temperature prediction method and system based on hybrid multilayer neural network model |
CN113192034A (en) * | 2021-04-30 | 2021-07-30 | 西安理工大学 | Mixed dye solution concentration detection method based on back propagation neural network |
CN113192034B (en) * | 2021-04-30 | 2024-02-02 | 西安理工大学 | Mixed dye solution concentration detection method based on counter-propagation neural network |
CN113435628A (en) * | 2021-05-28 | 2021-09-24 | 淮阴工学院 | Medium-and-long-term runoff prediction method and system based on linear discriminant analysis and IALO-ELM |
CN113435628B (en) * | 2021-05-28 | 2023-08-22 | 淮阴工学院 | Medium-long-term runoff prediction method and system based on linear discriminant analysis and IALO-ELM |
CN113494527A (en) * | 2021-07-30 | 2021-10-12 | 哈尔滨工业大学 | Constant force control method based on electromagnetic auxiliary type constant force spring support |
CN113494527B (en) * | 2021-07-30 | 2022-06-24 | 哈尔滨工业大学 | Constant force control method based on electromagnetic auxiliary type constant force spring support |
CN114994794A (en) * | 2022-06-24 | 2022-09-02 | 昆明学院 | Cloud particle phase state growth method for cloud cluster non-detection data area |
CN114994794B (en) * | 2022-06-24 | 2023-05-09 | 昆明学院 | Cloud particle phase growth method for cloud cluster non-detection data area |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
AU2020103741A4 (en) | Feed forward neural networks combined with extreme learning machine approach for large weather data | |
Chau et al. | A hybrid model coupled with singular spectrum analysis for daily rainfall prediction | |
Chen et al. | The strategy of building a flood forecast model by neuro‐fuzzy network | |
Paras et al. | A feature based neural network model for weather forecasting | |
Jalalkamali et al. | Monthly groundwater level prediction using ANN and neuro-fuzzy models: a case study on Kerman plain, Iran | |
Abrahamsen et al. | Machine learning in python for weather forecast based on freely available weather data | |
Farokhnia et al. | Application of global SST and SLP data for drought forecasting on Tehran plain using data mining and ANFIS techniques | |
Madhiarasan et al. | Analysis of artificial neural network: architecture, types, and forecasting applications | |
Kumar et al. | A time series ann approach for weather forecasting | |
Nagahamulla et al. | An ensemble of artificial neural networks in rainfall forecasting | |
Kadir et al. | Wheat yield prediction: Artificial neural network based approach | |
CN106709588A (en) | Prediction model construction method and equipment and real-time prediction method and equipment | |
Agrawal et al. | An application of time series analysis for weather forecasting | |
Al-Ghamdi et al. | Evaluation of artificial neural networks performance using various normalization methods for water demand forecasting | |
Zamani et al. | A comparative study on data mining techniques for rainfall prediction in Subang | |
Maqsood et al. | Weather forecasting models using ensembles of neural networks | |
Mantzari et al. | Solar radiation: Cloudiness forecasting using a soft computing approach. | |
Kumar et al. | Multilayer feed forward neural network to predict the speed of wind | |
Atiaa | Modeling of stage-discharge relationship for Gharraf River, southern Iraq by using data driven techniques: a case study | |
Soundararajan | A novel deep learning framework for rainfall prediction in weather forecasting | |
CN115062764B (en) | Intelligent illuminance adjustment and environmental parameter Internet of things big data system | |
Oyediran et al. | Performance evaluation of neural network MLP and ANFIS models for weather forecasting studies | |
Ghazali et al. | An application of Jordan Pi-sigma neural network for the prediction of temperature time series signal | |
Huisken | Soft-computing techniques applied to short-term traffic flow forecasting | |
de Lima et al. | Rainfall prediction for Manaus, Amazonas with artificial neural networks |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
FGI | Letters patent sealed or granted (innovation patent) | ||
MK22 | Patent ceased section 143a(d), or expired - non payment of renewal fee or expiry |