Three Intelligent Techniques For Weather Computer Science Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Artificial intelligence has developed rapidly. Techniques in artificial intelligence can be used to solve a lot of problems in completely different domains. I am endeavoring to predict weather forecasting using artificial intelligence techniques as a means to explore how to design, implement, and apply them to real world problems. Weather systems are extremely complex and poorly understood which makes weather prediction an excellent subject for them. The study is by comparing three artificial intelligence techniques such as artificial neural network, and genetic algorithm and fuzzy logic. These three techniques is very popular and the most successful AI techniques.

Keywords Artificial Neural Network, Genetic Algorithm, fuzzy logic, Weather Forecasting Prediction

1. Introduction

The idea of artificial Intelligence has been around for over half a century. Since the early days of computing there has always been the suggestion that one day a computing device would exhibit intelligence that would match or beat its human masters.

Artificial Intelligence (AI) is a multidisciplinary field whose goal is to automate activities that presently require human intelligence. Recent successes in AI include computerized medical diagnosticians and systems that automatically customize hardware to particular user requirements. (Chuck Williams 1983).

The purpose of this paper is to enlighten people and let them be aware of How Weather Forecasting can be predicted using an existing three AI techniques and make some comparison between them. The AI techniques that we going to use are fuzzy inference system, artificial neural network and genetic algorithm.

Weather forecasting is the application of science and technology to predict the state of the atmosphere for a given location. Human beings have attempted to predict the weather informally for millennia, and formally since the nineteenth century. Weather forecasts are made by collecting quantitative data about the current state of the atmosphere and using scientific understanding of atmospheric processes to project how the atmosphere will evolve.

A major part of modern weather forecasting is the severe weather alerts and advisories which the national weather services issue in the case that severe or hazardous weather is expected. This is done to protect life and property. Some of the most commonly known of severe weather advisories are the severe thunderstorm and tornado warning, as well as the severe thunderstorm and tornado watch. Other forms of these advisories include winter weather, high wind, flood, tropical cyclone, and fog. Severe weather advisories and alerts are broadcast through the media, including radio, using emergency systems as the Emergency Alert System which breaks into regular programming.

Weather Prediction is very important for some sectors and cannot do their work without prior weather information. Air Traffic is one example that relay on weather prediction information because the aviation industry is especially sensitive to the weather, accurate weather forecasting is essential. Fog or exceptionally low ceilings can prevent many aircraft from landing and taking off. Turbulence and icing are also significant in-flight hazards. Thunderstorms are a problem for all aircraft because of severe turbulence due to their updrafts and outflow boundaries, icing due to the heavy precipitation, as well as large hail, strong winds, and lightning, all of which can cause severe damage to an aircraft in flight.

There are a variety of end uses to weather forecasts. Weather warnings are important forecasts because they are used to protect life and property. Forecasts based on temperature and precipitation are important to agriculture, and therefore to traders within commodity markets. Temperature forecasts are used by utility companies to estimate demand over coming days. On an everyday basis, people use weather forecasts to determine what to wear on a given day. Since outdoor activities are severely curtailed by heavy rain, snow and the wind chill, forecasts can be used to plan activities around these events, and to plan ahead and survive them.

The simplest method of forecasting the weather, persistence, relies upon today's conditions to forecast the conditions tomorrow. This can be a valid way of forecasting the weather when it is steady state, such as during the summer season in the tropics. This method of forecasting strongly depends upon the presence of a stagnant weather pattern. It can be useful in both short range forecasts and long range forecasts.

2. Related Research

Many works have attempted to apply AI techniques for weather forecasting. C. M. Kishtawal, Sujit Basu, Falguni Patadia, and P. K. Thapliyal et al (2003) applied the feasibility of a nonlinear technique based on genetic algorithm for the prediction of summer rainfall over India. The genetic algorithm finds the equations that best describe the temporal variations of the seasonal rainfall over India, and therefore enables the forecasting of the future rainfall.

In research as evidenced by WONG Ka Yan, YIP Chi Lap and LI Ping Wah et al (2006) is to identify Weather systems such as tropical cyclones, fronts, troughs and ridges affect our daily lives and there are multiple atmospheric elements need to be considered, and the results may vary among forecasters. They contribute to the fields of pattern recognition and meteorological computing by designing a generic model of weather systems, along with a genetic algorithm-based framework for finding them from multidimensional numerical weather prediction data. It was found that their method not only can locate weather systems with 80% to 100% precision, but also discover features that could indicate the genesis or dissipation of such systems that could be ignored by forecasters.

Research conducted by Bjarne K. Hansen, Denis Riordan et al (2001) applied fuzzy k-nearest neighbors algorithm and concluded that querying a large database of weather observations for past weather cases similar to a present case that is designed and tuned with the help of a weather forecasting expert can increase the accuracy of predictions of cloud ceiling and visibility at an airport. Of significance to Case-Base Reasoning (CBR): They have shown how fuzzy logic can impart to CBR the perceptiveness and case discriminating ability of a domain expert. The fuzzy k-nn technique described in this thesis retrieves similar cases by emulating a domain expert who understands and interprets similar cases. The main contribution of fuzzy logic to CBR is that it enables us to use common words to directly acquire domain knowledge about feature salience. This knowledge enables us to retrieve a few most similar cases from a large temporal database, which in turn helps us to avoid the problems of case adaptation and case authoring.

Recently research conducted by N. Q. Hung, M. S. Babel, S. Weesakul, and N. K. Tripathi et al (2009) to improve rainfall forecast performance using an Artificial Neural Network (ANN). A real world case study was set up in Bangkok; 4 years of hourly data from 75 rain gauge stations in the area were used to develop the ANN model. The developed ANN model is being applied for real time rainfall forecasting and flood management in Bangkok, Thailand. Aimed at providing forecasts in a near real time schedule, different network types were tested with different kinds of input information. Preliminary tests showed that a generalized feed forward ANN model using hyperbolic tangent transfer function achieved the best generalization of rainfall. Especially, the use of a combination of meteorological parameters (relative humidity, air pressure, wet bulb temperature and cloudiness), the rainfall at the point of forecasting and rainfall at the surrounding stations, as an input data, advanced ANN model to apply with continuous data containing rainy and non-rainy period, allowed model to issue forecast at any moment. Additionally, forecasts by ANN model were compared to the convenient approach namely simple persistent method. Results show that ANN forecasts have superiority over the ones obtained by the persistent model. Rainfall forecasts for Bangkok from 1 to 3 h ahead were highly satisfactory. Sensitivity analysis indicated that the most important input parameter besides rainfall itself is the wet bulb temperature in forecasting rainfall.

In 2007 Mohsen Hayati and Zahra Mohebi et al applied the neural network models based on seasonal prediction for one day ahead temperature forecast for Kermanshah city, Iran; showed that Multi-Layer Perceptron (MLP) network witch trained and tested using ten years past (1996-2006) meteorological data has minimum error between exact and predicted value at each day and has a good performance, reasonable prediction accuracy and minimum prediction error in general. The forecasting reliability was evaluated by computing the mean absolute error between the exact and predicted value. The result showed that this network can be an important tool for temperature forecasting.

3. Comparative Analysis

3.1 Fuzzy Logic

Fuzzy logic suggests inaccuracy and imprecision. Webster¿½s dictionary defines the word fuzzy as ¿½not clear, distinct, or precise¿½ blurred.¿½ In a broad sense, fuzzy logic refers to fuzzy sets, which are sets with blurred boundaries, and in a narrow sense, fuzzy logic is a logical system that aims to formalize approximate reasoning.

Fuzzy logic is an approach to computer science that mimics the way a human brain thinks and solves problems. The idea of fuzzy logic is to approximate human decision-making using natural language terms instead of quantitative terms. It is formally defined as a form of knowledge representation suitable for notions that cannot be defined precisely, but which depend upon their context. It enables computerized devices to reason more like humans. Fuzzy logic technology has created a paradigm shift evident through many scientific and industrial applications.

The idea of fuzzy logic is to approximate human decision-making using natural language terms instead of quantitative terms. Fuzzy logic is similar to neural networks, and one can create behavioral systems with both methodologies. A good example is the use of fuzzy logic for automatic control: a set of rules or a table is constructed that specifies how an effect is to be achieved, provided input and the current system state. Using fuzzy arithmetic on uses a model and makes a subset of the system components fuzzy so that fuzzy arithmetic must be used when executing the model. In a broad sense, fuzzy logic refers to fuzzy sets, which are sets with blurred boundaries, and, in a narrow sense, fuzzy logic is a logical system that aims to formalize approximate reasoning. (Joseph Bih 2006)

3.1.1 Weather Forecasting using Case-Base Reasoning (CBR) and Fuzzy-Set:

CBR and fuzzy set theory each have their own unique well demonstrated strengths. So when both methods are combined in one system, the system stands the chance of inheriting the strengths of both methods. CBR is recommended to developers who are challenged to reduce the knowledge acquisition task, avoid repeating mistakes made in the past, reason in domains that have not been fully understood or modeled, learn over time, reason with incomplete or imprecise data and concepts, provide a means of explanation, and reflect human reasoning (Main et al. 2000).

Weather prediction presents special challenges for CBR. Weather is continuous, data-intensive, multidimensional, dynamic and chaotic. These properties make weather prediction a formidable proving ground for any CBR prediction system that depends on searching for similar sequences.

The weather prediction system (WIND-1) consists of two main parts, a large database of weather observations and a fuzzy k-nn algorithm, described as follows. The fuzzy k-nn algorithm measures the similarity between temporal cases, past and present intervals of weather observations. The algorithm is tuned by interviewing an experienced forecaster who describes various attribute difference thresholds that are to be used to signify various degrees of similarity (i.e., very near, near, slightly near). We design a similarity-measuring function, sim that is used to find k-nn for a present weather case and rank them according to their degree of similarity to the present weather. Given two cases, each identified by unique time indexes t1 and t2, sim returns a real number proportional to the degree of similarity of the two cases such that: 0.0 < sim(t1, t2) ? 1.0. Because all weather cases are unique and because the value of sim is calculated to double precision, sim can identify exactly k nearest neighbors. There are no null search results and no ties. The three steps to construct and use the algorithm are:

1. Configure similarity-measuring function.

2. Traverse case base to find k-nn.

3. Make prediction based on weighted median of k-nn.

Step 1 is performed only once and Steps 2 and 3 are performed every time a weather prediction is made. Step 1 is performed by interviewing a domain expert, and is thus the critical knowledge acquisition step in system design. For each continuous attribute, xi, the expert specifies thresholds for considering two such attributes to be very near, near, and slightly near each other (Figure 1).

Each experiment consists of a forecasting scenario. Five sets of experiments are conducted. In each set of experiments we systematically change the fixed parameters of WIND-1 and measure the resultant effects on forecast accuracy. The fixed parameters (independent variables) are: the attribute set, the number of analogs used to make forecasts, the size of the case base, and the fuzzy membership functions. The output (dependent variables) are, for each individual forecast, forecast values of cloud ceiling and visibility, and, for each set of experiments, a summary of the accuracy of all the forecasts made.

The first set of experiments varies the attribute set and shows that prediction accuracy increases as the number of attributes used for comparison increases. The second set of experiments varies k, the number of nearest neighbors that are used as the basis of predictions (k = 1, 2, 4, 8, ... , 256) and finds that maximum accuracy is achieved with k = 16. This suggests that WIND-1 is effective at identifying and ranking nearest neighbors, or, in meteorological terms, it finds the best analog ensemble. The third sets of experiments vary the size of the case base and shows that prediction accuracy increases as the size of the case base increases. The fourth and fifth sets of experiments pit WIND-1 using non-fuzzy sets against PC, and WIND-1 using fuzzy sets against PC, respectively. The non-fuzzy based prediction method is only slightly more accurate than PC, and fuzzy k-nn based prediction method is significantly more accurate than PC. The only variation between the two methods is the nature of the membership functions used to compare attributes of cases.

Based on our related work, experiments, and the results presented, it is concluded that querying a large database of weather observations for past weather cases similar to a present case using a fuzzy k-nearest neighbors algorithm that is designed and tuned with the help of a weather forecasting expert can increase the accuracy of predictions of cloud ceiling and visibility at an airport.

3.2 Genetic Algorithm (GA):

A genetic algorithm is an optimization heuristic that mimics natural processes, such as selection and mutation in natural evolution, to evolve solutions to problems whose solution spaces are impractical for traditional search techniques, such as branch-and-bound, or optimization techniques, such as linear programming. Since first described by Holland, genetic algorithms have been applied to a variety of learning and optimization.

A genetic algorithm typically begins with a random population of solutions (chromosomes) and, through a recombination process and mutation operations, gradually evolves the population toward an optimal solution. Obtaining an optimal solution is not guaranteed. The challenge is to design the process to maximize the probability of obtaining such a solution. The first step is the selection of the solutions in the current population that will serve as parents in the next generation of solutions. This selection requires that the solutions be evaluated for their fitness as parents: solutions that are closer to an optimal solution are judged higher, or more fit, than others.

In GA, search starts with an initial set of random solutions known as population. Each chromosome of population is evaluated using some measure of fitness function, which represents a measure of the success of the chromosome. Based on the value of the fitness functions, a set of chromosomes is selected for breeding. In order to simulate a new generation, genetic operators such as crossover and mutation are applied. According to the fitness value, parents and offspring¿½s are selected, while rejecting some of them so as to keep the population size constant for new generation. The cycle of evaluation selection-reproduction is continued until an optimal or a near-optimal solution is found.

Selection attempts to apply pressure upon the population in a manner similar to that of natural selection found in biological systems. Poorer performing individuals (evaluated by a fitness function) are weeded out and better performing, or fitter, individuals have a greater than average chance or promoting the information they contain to the next generation.

Crossover allows solutions to exchange information in a way similar to that used by a natural organism undergoing reproduction. This operator randomly chooses a locus and exchanges the subsequences before and after that locus between two chromosomes to create two offspring.

Mutation is used to randomly change (flip) the value of single bits within individual strings to keep diversity of a population and help a genetic algorithm to get out of a local optimum. It is typically used sparingly. (Arash Ghanbari et al 2010)

3.2.1 Weather Forecasting using Fuzzy-Set Genetic Algorithm:

Cities and towns were damaged by natural disasters that have violence cause significant damage and economic and social. If we can prevent disasters that may occur in advance is important. So an estimated rainfall data is important information for prevention disasters. The objective is to apply a fuzzy set theory to estimate rainfall. The genetic algorithm was applied to calibrate the fuzzy set model. The proposed model considered only a few basic hydrological parameters including temperature, humidity, wind speed and solar radiation. The proposed model was applied to estimate the rainfall in the Chi River Basin (in the northeast region of Thailand) using 5- minute historic data.

System inputs include the temperature, humidity, wind speed and solar radiation. Output is the rainfall data. There are four steps for developing fuzzy model as the following.

1. Transform the crisp inputs into fuzzy variable through the membership function, called Fuzzification process. The number and type of membership functions are constructed based on statistical data and experience of engineers, generally upon the considering problem. This study used the triangular membership function for describing the input and output variables.

2. The fuzzy rule bases are created using the 5-min historical data of all parameters and fuzzy operator. The historical data were collected from the Kosumphisai Meteorological Station. These fuzzy operators AND and OR are applied to combine the input variables.

3. Apply the input membership functions and the rule bases to obtain the output membership functions. This step is done by the implication method which obtaining a fuzzy set of output when given a single number of each inputs. Then the output membership functions of each rule are jointed to one output fuzzy set, called aggregation process.

4. The process is defuzzification that a fuzzy set of output is converted into a single crisp value. The most common defuzzification method is the ¿½centroid¿½ evaluation, which returns the center of area.

The adequacy of the fuzzy model is evaluated by considering the coefficient of determination (R2) which defined based on the irrigation efficiency estimation errors as:

Where, ?j is the estimated rainfall of the 5 min duration record j which calculated using fuzzy model, is the actual rainfall of 5 min duration record j which collected from the Meteorological Station, and are respectively the average of above mentions and m is the number of data record. The fuzzy model is calibrated by adjusting the membership functions and rule bases using the genetic algorithms technique, these performances will be stopped when the results obtained the highest coefficient of determination (closed to 1.0).

GA requires encoding schemes that transform the decision variables into chromosome. Then, the genetic operations (reproduction, crossover and mutation) are performed. These genetic operations will generate new sets of chromosomes. In this study, each decision variable represents a parameter of membership function. The objective function of the search is to maximize the coefficient of determination (R2). This study used population size = 80, crossover probability = 0.9and mutation probability = 0.01. There are 4 cases that were considered in the study including 2-input parameter, 3-input parameter, 4-input parameter and 4-input parameter with duration of rainfall.

2-input parameter consideration: The input parameters were applied to create the fuzzy model only 2 variables. However, all couples of input variables were used to create and test with their historical data record. Table 1 shows coefficient of determination of each couple. The couple of relative humidity and solar radiation provide the highest coefficient of determination as 0.2968. Table 2 showed the lag time providing the highest coefficient of determination for 2-input consideration because rainfall event may affect from climate and hydrologic variables in the past according to the previous study (Frencha et al., 1992), hence the lag time by 5, 10, 30 min, 1-3, 9, 12, 18 and 24 h were performed. They indicated that the highest coefficients of determination of the lag time consideration are higher than the coefficient of determination of the present time. Figure 2 present the coefficient of determination of lag time consideration for 2-input parameter (relative humidity-solar radiation). They present that the coefficient of determination of lag time from 5 min-9 h are not significantly deferent. It can conclude that rainfall event is affected by the previous situation more than the present situation (Frencha et al., 1992).

3-input parameter consideration: Table 3 shows the coefficient of determination of 3-input parameter. The input of temperature-relative humidity-solar radiation provided the highest coefficient of determination as 0.3359. Lag time conditions were performed; the coefficients of determination for all cases were present in Table 4. Figure 3 present the coefficient of determination of lag time consideration for 3-input parameter (Temperature-Relative Humidity-Solar Radiation). The coefficient of determination of lag time 2 h is the highest.

The result indicated that input of temperature-relative humidity-solar radiation provided the highest coefficient. When considered by any lag time, they found that the coefficients of determination are increase for all cases (Table 4). As shown in Fig. 3, the coefficient of determination of lag time 2 h is the highest. However, the coefficients of determination of lag time over 2 h are slightly decrease.

4-input parameter consideration: The 4-input parameters were applied to create the fuzzy model using the present time of data record. The highest coefficient of determination is 0.3562. When considered by any lag time, it found that the coefficient of determination is increase to 0.8383 at lag time 1 h. Figure 4 presents the coefficient of determination of lag time consideration for 4-input parameter. The result indicated that the coefficients of determination of lag time from 5 minute to 24 h are similarly because all parameters are together effect to rainfall.

3.3 Artificial Neural Network (ANN)

The term neural network was traditionally used to refer to a network or circuit of biological neurons. The modern usage of the term often refers to artificial neural networks, which are composed of artificial neurons or nodes. Artificial neural networks are composed of interconnecting artificial neurons (programming constructs that mimic the properties of biological neurons). Artificial neural networks may be used for solving artificial intelligence problems without necessarily creating a model of a real biological system.

One type of network sees the nodes as ¿½artificial neurons¿½. These are called artificial neural networks (ANNs). An artificial neuron is computational model inspired in the natural neurons. Natural neurons receive signal through synapses on the dendrites or membrane of the neuron. When the signals received are strong enough (surpass certain threshold), the neuron is activated and emits a signal though the axon. This signal might be sent to another synapse, and might activate other neurons.

The complexity of real neurons is highly abstracted when modeling artificial neurons. These basically consist of inputs (like synapse), which are multiplied by weights (strength of the respective signals), and then computed by a mathematical function which determines the activation of the neuron. Another function (which may be identify) computes the output of the artificial neuron (sometimes in dependence of a certain threshold). ANNs combine artificial neurons in order to process information.

The higher a weight of an artificial neuron is, the stronger the input which is multiplied by it will be. Weights can also be negative, so we can say that the signal is inhibited by the negative weight. Depending on the weights, the computation of the neuron will be different. By adjusting the weights of an artificial neuron we can obtain the output we want for specific inputs. But when we have an ANN of hundreds or thousands of neurons, it would be quite complicated to find by hand all the necessary weights. But we can find algorithms which can adjust the weights of the ANN in order to obtain the desired output from the network. This process of adjusting the weights is called learning or training.

Since the function of ANNs is to process information, they are used mainly in fields related with it. There are wide variety of ANNs that are used to model real neural networks, and study behavior and control in animals and machines, but also there are ANNs which are used for engineering purposes, such as pattern recognition, forecasting, and data compression. (Carlos Gershenson 2001)

3.3.1 Weather Forecasting using Radial Basis Function (RBF) Neural Network:

A radial basis function network is an artificial neural network that uses radial basis functions as activation functions. It is a linear combination of radial basis functions. The configuration of the neural network depends highly on the problem. Therefore, it is left with the designer to choose an appropriate number of input, output and hidden layer nodes based on experience. Thus, an appropriate architecture is determined for each application using the trial and error method. The learning rate parameter and momentum term were adjusted intermittently to speed up the convergence.

In RBF networks, determination of the number of neurons in the hidden layer is very important because it affects the network complexity and the generalizing capability of the network. In the hidden layer, each neuron has an activation function. The Gaussian function, which has a spread parameter that controls the behavior of the function, is the most preferred activation function (Ajeel 2010). The training procedure of RBF networks also includes the optimization of spread parameters of each neuron. Afterwards, the weights between the hidden layer and the output layer must be selected appropriately. Finally, the bias values which are added with each output are determined in the RBF network training procedure. RBF network is a type of feed forward neural network composed of three layers, namely the input layer, the hidden layer and the output layer. A general block diagram of an RBF network is illustrated in Fig. 7.

A RBF network with m outputs and hidden nodes can be expressed as:

Considering this argument, the RBF network with additional linear input connections is used. The proposed network allows the network inputs to be connected directly to the output node via weighted connections to form a linear model in parallel with the non-linear standard RBF model

(Ghods and Kalantar, 2010). The new RBF network with m outputs, n inputs, hidden nodes and n1 linear input connections can be expressed as:

where the ?¿½s and vl¿½s are the weights and the input vector for the linear connections may consist of past inputs and outputs. The ?¿½s can be estimated using the same algorithm. As the additional linear connections introduce a linear model, no significant computational load is added to the standard RBF network training (Gorriz et al., 2004). Furthermore, the numbers of required linear connections are normally much smaller than the number of hidden nodes in the RBF network. In the present study, given least squares algorithm with additional linear input connection features is used to estimate weight.

BPN and RBF are trained with sample of six hundred patterns. The performance of the RBF is compared with performance of the BPN for weather forecasting. The training of RBF is faster compared that of BPN. The classification to predict rainfall is enhanced by RBN which is depicted in Table 1. Performance of classification of rainfall prediction by BPN and RBF is shown in Fig. 8 and 9.

4. Discussion

In this paper I present several artificial intelligence techniques to predict weather and use different data on each technique, so I can see the best technique here. The first technique was Genetic Algorithm to estimate rainfall. Genetic algorithm (GA) was applied to find the equation that best fits the rainfall data in one part of the dataset, the training set. The obtained rainfalls of the improved model are close to the rainfall of the actual rainfall record. Furthermore, the results presented that the genetic algorithm calibration provided the optimal condition of membership function. In addition, there are numerous advantages of using of GA, such as not depending on analytical knowledge, robustness, and intuitive operation which are in the weather domain. All of these characteristics have made GAs strong candidates weather forecasting. However, there are also several disadvantages of using GAs in weather forecasting research that have made researchers turn to other search techniques, such as Probabilistic, Expensive in computational resources, Prone to premature convergence and Dif?cult to encode a problem in the form of a chromosome.

The second technique was Fuzzy Algorithm; it was found that querying a large database of weather observations for past weather cases similar to a present case using a fuzzy similarity measure that is designed and tuned with the help of a weather forecasting expert together with a k-nearest neighbors algorithm and weighted adaptation can increase the accuracy of predictions of cloud ceiling and visibility at an airport. The main contribution of fuzzy logic is that it enables us to use common words to directly acquire domain knowledge about feature salience. This knowledge enables us to retrieve a few most similar cases from a large temporal database, which in turn helps us to avoid the problems of case adaptation and case authoring. However, Fuzzy Logic need to be crafted by hand, May not scale well to large rule sets and its performance can¿½t be easily predicted.

The third technique was Artificial Neural Network; it gained great popularity in weather prediction because of their simplicity and robustness. Plus, it can learn more complicated class boundaries, fast application and can handle large number of features. The performance of back propagation neural network (BPN) and radical basis functioned neural network (RBF) is compared. Back propagation algorithm is time consuming and the performance is heavily dependent on the network parameters. Compared to BPN, RBF gives the overall best results in terms of accuracy and fastest training time. RBFN are much faster and more reliable for the weather forecasting. These proportions make it more effective for fast real time weather forecasting. However, its training time is slow and hard to interpret plus it¿½s hard to implement trial and error for choosing number of nodes.

5. Conclusion

To sum up, weather forecasting is the application of science and technology to predict the state of the atmosphere for a given location. Human beings have attempted to predict the weather informally for millennia, and formally since the nineteenth century. Weather forecasts are made by collecting quantitative data about the current state of the atmosphere and using scientific understanding of atmospheric processes to project how the atmosphere will evolve.

In this paper we studied how to predict weather using three artificial intelligence approach: artificial neural network, genetic algorithm, and fuzzy logic. Some of the sample study shows that we also can use more than one technique to process weather forecasting, and we call it hybrid like Fuzzy-Set Genetic Algorithm subtitle that we explain. However the three techniques show that algorithm is working and can forecast weather very well. There are many other factors that influence forecasting result.



I¿½d like to sincerely thank almighty Allah for all his grants that he bestowed on me. I truly acknowledge the valuable time, patience, support of my supervisor Prof. Dr. AZURALIZA BINTI ABU BAKAR for the valuable guidance and advice.