25
CHAPTER 2 REVIEW OF RELATED WORKS 2.1 INTRODUCTION Time series plays a vital role in the branches of Stock prices, Global surface temperature, Economy Population growth and agricultural production. Time series is the sequence of observations ordered in time. Mostly these observations are collected at equally spaced and discrete time interval. The basic assumption of time series analysis is some aspects of the past pattern which will continue to remain in the future. In recent years, many researchers have used fuzzy time series to handle forecasting problems. In this thesis, Markov modeling is a major statistical tool used to predict the fuzzy time series data. The advantage of using Markov modeling gives better forecasting accuracy for predicting time series data. Many works are related to fuzzy time series in the applications of stock price prediction, agriculture commodities prediction, and Economy and university enrollment data. The works related to Markov modeling fuzzy time series can be discussed.

Chapter 2 - shodhganga.inflibnet.ac.inshodhganga.inflibnet.ac.in/bitstream/10603/26799/8/08_chapter 2.pdf · fuzzy time series to handle forecasting problems. In this thesis, Markov

  • Upload
    others

  • View
    15

  • Download
    0

Embed Size (px)

Citation preview

  • CHAPTER 2

    REVIEW OF RELATED WORKS 2.1 INTRODUCTION

    Time series plays a vital role in the branches of Stock prices,

    Global surface temperature, Economy Population growth and

    agricultural production. Time series is the sequence of observations

    ordered in time. Mostly these observations are collected at equally

    spaced and discrete time interval. The basic assumption of time series

    analysis is some aspects of the past pattern which will continue to

    remain in the future. In recent years, many researchers have used

    fuzzy time series to handle forecasting problems.

    In this thesis, Markov modeling is a major statistical tool used

    to predict the fuzzy time series data. The advantage of using Markov

    modeling gives better forecasting accuracy for predicting time series

    data. Many works are related to fuzzy time series in the applications of

    stock price prediction, agriculture commodities prediction, and

    Economy and university enrollment data. The works related to Markov

    modeling fuzzy time series can be discussed.

  • 24

    2.2 MARKOV MODEL FUZZY TIME SERIES (MM-FTS)

    Fuzzy Time Series (FTS) is introduced by Song and

    Chissom [1993a]. Several fuzzy time series model is developed in the

    literature during the last two decades. Zadeh [1965] discussed fuzzy

    sets and it is characterized by a membership function which assigns

    the value to each object grade membership ranging from zero to one.

    Song and Chissom [1993b] have initiated the study of time

    invariant and time variant forecasting models using fuzzy time series

    for enrollment data of University of Alabama. Also compare the

    predicted values of fuzzy time series with those of non linear

    regression models.

    Song et al., [1993a] have used time-invariant fuzzy time series

    model, while Song and Chissom [1994] used time-variant model for

    the same problem. In their example, the crisp data is fuzzified into

    linguistic values to illustrate the fuzzy time series method using fuzzy

    set theory and its modeling by applying fuzzy relations equations and

    approximate reasoning.

    Sullivan et al., [1994] reviewed the first-order time-variant fuzzy

    time series model and the first-order time-invariant fuzzy time series

    model presented by Song and Chissom [1993b], where their models

    are compared with each other and with a time-variant Markov model

    using linguistic labels with probability distributions. The results of

  • 25

    this model enrollment data are compared with three traditional time

    series models such as a first order auto regressive model and second

    order auto regressive models. However, the Markov model has an

    advantage that any repeated transitions in the data are taken into the

    account of estimating the model parameters, whereas the fuzzy time

    series model based on the usual max-min operations cannot perform

    the same.

    Song et al., [1994] proposed an approach for developing the

    time-variant fuzzy time series models with an example which

    presented the process of forecasting the enrollment for the University

    of Alabama. To defuzzify the output of the model, a 3-layer back

    propagation neural network was trained and used as the defuzzifier.

    Among the three different defuzzification methods, neural network

    method yielded the best result.

    Song et al., [1995] have constructed fuzzy time-series model by

    means of defining some new operations on fuzzy numbers.

    Chen [1996] developed a simplified method for time series forecasting

    using the arithmetic operations rather than using complicated

    maximum – minimum composition operations. The proposed method

    not only gives good forecast for the University enrollments, but also

    gives robust forecast even when the historical data are not accurate.

  • 26

    Ranaweera et al., [1996] proposed a fuzzy logic model for short

    term load forecasting which was applied to historical weather and load

    data. Since Conventional fuzzy systems cannot operate with random

    phenomena, the Control processes in real-life plants were used.

    They have deal the stochastic processes as Markov modeling

    approach.

    Song et al., [1997] have discussed fuzzy stochastic fuzzy time

    series and three different models. Ping-Teng Chang [1997] has

    developed a fuzzy technique for trends and seasonality forecasting

    through fuzzy regression and fuzzy trend of time-series model. Fuzzy

    forecasting and the analysis of the fuzziness of the forecasts are the

    features of this method which differs from the traditional forecasting

    techniques for seasonality.

    Jeng- Ren Hwang et al., [1998] have proposed new method for

    handling forecasting problems based on time variant fuzzy time series.

    In the proposed method where the variation of enrollment of the

    current year is related to the trend of the enrollments of past years.

    Song [1999] has developed a seasonal forecasting model for fuzzy time

    series with minor modifications of fuzzy logical relationship and the

    membership functions. The model can be used directly for seasonality

    forecast.

  • 27

    Sullivan and Woodall [1999] have discussed three methods for

    estimating Markov transition matrices when observed state

    probabilities are not all either zeros or ones and a simulation-based

    comparison of the performance of the estimators.

    Fang-Mei Tseng et al., [1999] have proposed the method of fuzzy

    seasonal time series for forecasting the production value of the

    mechanical industry in Taiwan. This method provided the decision

    makers with insight regarding the possible future situation. Chen et

    al., [2000] have discussed fuzzy time series method and adapted to

    temperature data in Taiwan.

    Nguyen et al.,[2002] has developed the mathematical modeling

    of domains of linguistic variables, which gives a unified algebraic

    approach to the natural structure of domains of linguistic variables.

    Tapio Frantti et al., [2001] have obtained an anticipatory Fuzzy Logic

    Advisory Tool (FLAT) for predicting the demand for signal transmission

    device. The design principle for a general predictor is also explained.

    The algorithm is implemented as a practical computer program and it

    is applied to real manufacturing data.

    Fang-Mei Tsenga et al., [2001] have developed Fuzzy ARIMA

    (FARIMA) model and applied it for forecasting the exchange rate of NT

    dollars to US dollars. With the help of this model it is possible for

    decision makers to forecast the best and worst possible situations

    based on fewer observations than the ARIMA model.

  • 28

    Fang-Mei Tsenga et al., [2002] have proposed fuzzy seasonal ARIMA

    (FSARIMA) forecasting model. It is used to forecast two seasonal time

    series data of the total production value of the Taiwan machinery

    industry and the soft drink time series data.

    Song [2003] has proposed the sample autocorrelation functions

    of fuzzy time series and used in model selection. The main idea is to

    select a number of different data sets from each fuzzy set and

    calculate the sample auto correlation function for each data set. Three

    different auto correlation functions are proposed and examples are

    also presented. Javier Contreras et al., [2003] have proposed a method

    to predict next-day electricity prices based on ARIMA methodology.

    Yimin Xiong et al., [2004] have proposed a model-based method for

    clustering univariate and simple multivariate ARMA time series.

    Hsuan-Shih Lee et al., [2004] have presented an improved

    method to forecast university enrollments based on the fuzzy time

    series. The method proposed not only defines the supports of the fuzzy

    numbers that represent the linguistic values of the linguistic variable

    more appropriately, but also makes the RMSE smaller.

    Melike Sah et al., [2005] have proposed a method to attain better

    forecasting accuracy using time-invariant fuzzy time series. It should

    be emphasized that it uses only historical data in the numerical form

  • 29

    without any addition pieces of knowledge for forecasting. In addition

    the method is tested on different number of fuzzy sets for the purpose

    of examining the forecasting accuracy.

    Ruey-Chyn Tsaur et al., [2005] have proposed fuzzy relation

    matrix affecting the forecasting performance and proposed an

    arithmetic procedure for deriving fuzzy relation matrix method using

    Fuzzy relation analysis in fuzzy time series. Fuzzy relation is a crucial

    connector in presenting fuzzy time series model. Also the concept of

    entropy is applied to measure the degrees of fuzziness when a time

    invariant matrix is derived.

    Hui-Kuang Yu [2005] has proposed weighted models to tackle

    two issues in fuzzy time series forecasting, viz., recurrence and

    weighting. It is argued that recurrent fuzzy relationships, which are

    simply ignored in previous studies, should be considered in

    forecasting. It also recommended that different weights shall be

    assigned to various fuzzy relationships. The weighted model is

    compared with local regression models in which weight function plays

    an important role in forecasting.

    Ping-Feng Pai et al., [2005] have proposed autoregressive

    integrated moving average (ARIMA) model which has become one of

    the most widely used linear models in time series forecasting.

    Nai-Yi-Wang et al., [2009] have presented a new method to predict the

  • 30

    temperature and the Taiwan Futures Exchange (TAIFEX), based on

    automatic clustering techniques and two-factor high-order fuzzy time

    series. Tasha Inniss [2006] has applied this technique to weather and

    aviation data to determine probabilistic distributions of arrival

    capacity scenarios, which can be used for seasonal forecasting and

    planning.

    Alonsoa et al., [2006] have proposed time series method based

    on the probability density of the forecasts using nonparametric kernel

    estimator. Hao-Tien Liu [2007] has proposed improved time-variant

    fuzzy time series method. The proposed method takes into

    consideration of Window base, length of interval, degrees of

    membership values, and existence of outliers. The improved method

    provides decision makers with more precise forecasted values.

    Sheng-Tun Li [2007] has proposed a deterministic forecasting

    model to manage the crucial issues. In addition, an important

    parameter, the maximum length of subsequence in a fuzzy time series

    resulting in a certain state, is deterministically quantified.

    Hao-Tien Liu [2009] has proposed a method to design an improved

    fuzzy time series forecasting method in which the forecasted value

    would be a trapezoidal fuzzy number instead of a single point value.

    Also, the decision analyst can gather information about the possible

    forecasted ranges under different degree of confidence.

  • 31

    Guoqi Qian et al., [2007] have proposed a feasible computing

    method based on the Gibbs sampler. By this method, model selection

    is performed through a random sample generation algorithm and also

    a model of fixed dimension in which the parameter estimation is done

    through the maximum likelihood method.

    Hye-young Jung et al., [2008] have proposed rearranged interval

    method to reflect the fluctuation of historical data and to improve the

    forecasting accuracy of fuzzy time series. This is discussed and the

    forecasting accuracy is evaluated. Empirical analysis shows that this

    method in forecasting accuracy is superior to existing methods and it

    fully reflects the fluctuation of historical data.

    Ashraf Abd-Elaal et al.,[2010] have proposed a fuzzy time series

    technique based on fuzzy c- mean clustering. This method is adapted

    to university enrollment data and it is improved the forecasting

    accuracy compared with existing models.

    Srivastava [2011] has presented a new method which gives

    short term forecast of agricultural production based on fuzzy time

    series. The study used fuzzy set theory and applied in fuzzy time

    series model and artificial neural network to forecast the production of

    food grains. Pushparani Suri et al., [2011] have presented a simple,

    systematic and iterative methodology for forecasting gold price. This

    method is realized on fuzzy clustering and weighted least square.

  • 32

    Hemant Kumar [2011] has proposed relation matrix method and

    it is used to reduce its calculation. This method is most

    remarkable one which gives the most plausible range of forecasting

    and it is in the form of interval rather than a single value.

    Seyed Morteza Hosseini et al., [2011] have demonstrated the results of

    test data from the currency market shows that this combined

    approach have to be more careful from existing fuzzy time series

    model. The practical results obtained show that this model have a

    good application.

    2.3 HIDDEN MARKOV MODEL FUZZY TIME SERIES (HMM – FTS)

    Rabiner et al., [1986] have proposed the theory and

    implementation of Markov modeling technique and applied it to

    speech recognition problems. Rabiner [1989] has developed theory of

    hidden Markov models from the simplest concepts to the most

    sophisticated models and focused on physical explanations of the

    basic mathematics. Rabiner also illustrated the applications of the

    theory of HMM’s to simple problems in speech recognition and pointed

    out how the techniques could be applied to more advanced speech

    recognition problems.

    Juang et al., [1991] have reviewed the statistical method of

    HMM's. It reveals that the strengths of the method lie in the consistent

    statistical framework that is flexible and versatile, particularly for

  • 33

    speech applications, and the ease of implementation that makes the

    method practically attractive. It is pointed out some areas of the

    general HMM method that deserves more attention with the hope that

    increased understanding will lead to performance improvements for

    many applications.

    Chen et al., [2000] have proposed two factor time variant

    algorithms based on fuzzy time series model and applied it to forecast

    the daily temperature. Both the algorithms have the advantage of

    obtaining good forecasting results.

    Rafiul Hassan et al., [2006] have proposed a single Hidden

    Markov Model based on clustering method, which identifies a suitable

    number of clusters in a given dataset without using prior knowledge

    about the number of clusters. Initially, the dataset is partitioned into

    windows of fixed size based on the HMM log likelihood values. This

    model provides a framework for identifying the most appropriate

    number of clusters (windows of varying sizes). After determining the

    number of clusters, the data values are then labeled and allocated to

    clusters.

    Ching-Hsue Cheng et al., [2007] have discussed the fuzzy time

    series method based on rough set theory and this method is applied to

    stock price index forecasting problem. Sheng - Tun Li [2008] has

    studied fuzzy time series which has increasingly

  • 34

    attracted much attention due to its salient capabilities of tackling

    uncertainty and vagueness inherent in the data collected.

    Ching-Hsue Cheng et al., [2008] have introduced a new multiple-

    attribute fuzzy time series method based on fuzzy clustering

    technique. The methods of fuzzy clustering are integrated in the

    processes of fuzzy time series to partition datasets objectively and

    enable processing of multiple attributes.

    Sheng-Tun Li et al., [2009] have proposed a new forecasting

    model based on Hidden Markov Model for fuzzy time series to realize

    the probabilistic state transition and also conducted experiments of

    forecasting a real-world temperature application to validate the better

    accuracy of the proposed model achieved over traditional fuzzy time

    series models.

    Rafiul Hassan [2009] has presented a novel combination of

    hidden Markov model and fuzzy models for forecasting stock market

    data. The model is tested by preparing forecasts for the financial time

    series data of six stock prices.

    Ming Dong et al., [2009] have proposed the states of hidden

    semi Markov models are used to represent the PM2.5 concentration

    levels. The model parameters are estimated through modified forward

    backward training algorithm. It can be used to predict PM2.5

    concentration levels. Jens Runi Poulson [2009] has been developed

  • 35

    new fuzzy time series method which combines aggregation and

    particle swarm optimization techniques. By combining these

    techniques, forecast rule can be individually tuned to match the data.

    Sheng-Tun Li et al., [2010] have proposed a novel forecasting

    model based on the Hidden Markov model by enhancing Sullivan and

    Woodall’s work to allow handling of two-factor forecasting problems.

    It is built upon an HMM in which the fuzzy relationships are

    formulated as state transitions. So that it can handle two factor

    forecasting problems. Moreover, in order to make a nature of

    conjecture and randomness of forecasting more realistic, the Monte

    Carlo method is used to estimate the outcome.

    Edmundo de Souza e Silva et al., [2010] have investigated the

    performance of a hidden Markov model in predicting future crude oil

    price movements. Additionally, they developed forecasting

    methodologies consists of three steps. First, they employed wavelet

    analysis to remove high frequency price movements, and then the

    hidden Markov model is used to forecast the probability distribution of

    the price return accumulated over the next F days.

    Suresh et al., [2011] have analyzed Global surface temperature

    with CO2 data using hidden Markov model in fuzzy time series. This

    method is used to forecast successive year’s of global surface

    temperature data.

  • 36

    2.4 HIGHER ORDER MARKOV MODEL FUZZY TIME SERIES

    (HOMM – FTS)

    Adrian Raftery et al., [1994] have introduced the mixture

    transition distribution model for higher order Markov chains. Also

    proposed a computational algorithm for maximum likelihood

    estimation which is based on a way of reducing the large number of

    constrains.

    Chen [2002] has presented present a new method for handling

    forecasting problem using a high-order fuzzy time series model, where

    the historical enrollments of the University of Alabama are used to

    illustrate the forecasting process.

    Vasek Chavatal [1983] has discussed linear programming

    problem formulated from optimization problem and solved by simplex

    method. Adrian Raftery [1985] has introduced higher order Markov

    chain model which combines realism and parsimony.

    Ching et al., [2002] have proposed a multivariate Markov chain

    model for modeling multiple categorical data sequences and also

    developed efficient estimation methods for the model parameters.

    Ching et al., [2003] have proposed a new higher-order Markov model

    whose number of states is linear and also developed a new parameter

    estimation method based on linear programming.

  • 37

    Tie Liu [2010] has adopted Markov chains model to analyze and

    predict the time series. Some series can be expressed by a first-order

    discrete-time Markov chain and others must be expressed by a higher-

    order Markov chain model. Ching et al.,[2008] has proposed an nth-

    order multivariate Markov chain model for modeling multiple

    categorical data sequences such that the total number of parameters

    are of O(ns2m2).

    Yutong Li et al., [2008] have presented a stochastic simulation

    approach to synthetic series of weather data to evaluate the

    performance of open cycle solar desiccant air condition system of

    Hong Kong. The results reveal that the open cycle desiccant system

    can meet most of the latent load through out the cooling season if the

    components are proper sized and energy savings.

    Zhilong Wang et al., [2009] have presented a higher-order

    multivariate Markov chain model combined with particle swarm

    optimization algorithm, capable of searching for the optimal parameter

    values η for level characteristics value to obtain a high accuracy model

    for forecasting of multidimensional time series. Particle swarm

    optimization algorithm is employed to optimize the coefficient of level

    characteristics value.

    Ersoy Oz [2011] has discussed the application of monthly

    changes of the US Dollar selling rates and the monthly changes of the

  • 38

    Euro selling rates. The two changes are taken into consideration as

    two categorical data sequences and it is revealed with multivariate

    Markov chain model to what degree these sequences affect each other

    in Turkey.

    Chi-Chen Wang et al., [2011a] have compared the applications

    of the forecasting methods Auto Regressive Integrated Moving Average

    (ARIMA) time series model and fuzzy time series model by heuristic

    models on the amount of export values in Taiwan.

    Chi-Chen Wang [2011b] has attempted to use information of

    export values in Taiwan as an example to test whether the fuzzy time

    series is indeed practical in its forecast of macro economic variables.

    By comparing fuzzy time series with ARIMA time series method, better

    understanding of the appropriateness of forecasting model could be

    obtained.

    Hongxing Yang et al., [2011] have introduced a new method to

    generate the annual weather data by using the first order multivariate

    Markov chain model. The weather variables are described in a

    stochastic way and multiple categorical sequences are generated by

    similar source.

  • 39

    2.5 HIGHER ORDER COMPUTATIONAL FUZZY TIME SERIES

    Huarng [2001a] has proposed an effective length of intervals to

    improve the forecasting accuracy and introduced two methods such as

    average based and distribution based methods. Distribution based

    length is the largest length which is smaller than at least half of the

    first differences of data. Average based length is set to one half of the

    average of the first differences of data. Both can be used as effective

    lengths to improve forecasting in fuzzy time series.

    Huarng [2001b] has demonstrated how the heuristic model

    outperforms of Chen’s and other models using the enrollment

    forecasting at the University of Alabama. This study assumes that

    there is heuristic knowledge showing the increase or decrease in

    enrollment for the next year. With this heuristic, the heuristic model

    forecasts the enrollment better than the other models. In the heuristic

    models, the heuristic knowledge is used to guide the search for

    suitable fuzzy sets for index forecasting.

    Chen [2002] has presented a forecasting method based on high-

    order fuzzy time series. From the proposed model he developed an

    algorithm to forecast the enrollments of the University of Alabama,

    where the historical enrollment data at the University of Alabama are

    used to illustrate the forecasting process.

  • 40

    Chen et al., [2004] has proposed fuzzy time series belonging to

    first order and time–variant methods. It is possible to get higher

    forecasting accuracy rate for forecasting enrollments than the existing

    methods.

    Sheng –Tun -Li et al., [2004] have presented a new approach to

    handling the issue by applying the natural partitioning technique,

    which can recursively partitioned the universe of discourse level by

    level in a natural way. Experimental results found that the enrollment

    data of the University of Alabama results can forecast the data

    effectively and efficiently and outperforms the existing models.

    Chung-Ming Own et al., [2005] have proposed an enhanced

    fuzzy time series model, called heuristic high-order fuzzy time series

    model, to deal forecasting problems. The proposed method eliminates

    ambiguities at forecasting and requires a vast memory for data

    storage. The empirical analysis reveals that the proposed method yield

    more accurate forecasts. Moreover, the forecasting model can be

    restricted in the acceptable-order fuzzy time series to reduce the

    memory needed for the data storage.

    Huarng et al., [2006] have suggested that ratios, instead of

    equal lengths of intervals, including distribution based length and

    average based length can more properly represent the intervals among

  • 41

    observations. The method also suggests that ratio-based lengths of

    interval can be applied to improve fuzzy time series forecasting.

    Ching-Hsue Cheng et al., [2006] have proposed two approaches

    to improve the persuasiveness in determining the universe of

    discourse, length of intervals and membership functions of fuzzy time

    series. The first approach is using Minimize entropy principle

    approach (MEPA) to partition the universe of discourse and build

    membership functions, and the second is using Trapezoid fuzzification

    approach (TFA).

    Li-Wei Lee et al., [2006] have proposed a new method for

    forecasting temperature and TAIEX, based on two factor high order

    fuzzy time series. The two factor high-order fuzzy logical relationship

    is used to increase the forecasting accuracy rate of prediction.

    Shiva Raj Singh [2007] proposed a simple time variant method

    for fuzzy time series forecasting. This method overcomes the deficiency

    of ambiguity in trends of data and also does not need the heuristic

    function. It provides simple computational algorithms for complexity

    in linear order. It minimizes the time of generating relation equations

    by using min-max composition operations and the time consumed by

    the various defuzzification processes to be applied for getting crisp

    forecast. The proposed algorithm is implemented for forecasting

  • 42

    enrollments of university of Alabama and the results are compared

    with the existing methods to show its superiority.

    Tahseen Ahmed Jilani et al., [2008d] have presented new fuzzy

    metrics for high order multivariate fuzzy time series forecasting for car

    road accident casualties in Belgium. Sun Xihao et al. [2008] have

    proposed Average – based fuzzy time series model, which can be used

    to adjust the lengths of intervals determined during the early stages of

    forecasting and that method is applied to daily stock index in

    shanghai.

    Tahseen Ahmed Jilani et al., [2008b] have proposed a new

    method for time series forecasting, having simple computational

    algorithm of complexity of linear order. The method first predicts the

    trend of the future value and then use the quantile based fuzzy

    forecasting approach.

    Tahseen Ahmed Jilani et al., [2008a] have presented two new

    multivariate fuzzy time series forecasting methods. These methods

    assumed m-factors with one main factor of interest. Stochastic fuzzy

    dependence of order k is assumed to define general methods of

    multivariate fuzzy time series forecasting and control. This method

    provided a general work for forecasting, that can be increased by

    increasing the stochastic fuzzy dependence and the simplicity of

    computation used triangular membership function.

  • 43

    Tahseen Ahmed Jilani et al., [2008c] have proposed other

    method based on frequency density partitioning applied for historical

    enrollments of Alabama. The proposed method uses heuristic

    approach to define frequency-density –based partitions of the universe

    of discourse and also uses a trend predictor to calculate the forecasted

    value. The trend predictor is used to adjust the weights of the

    proposed fuzzy metric for forecasting. The method is found to be

    robust and can handle the problem of inaccuracy in the data set.

    Shiva Raj Singh [2008] has proposed a method which minimizes

    the complicated computations of fuzzy relational matrices and a

    suitable defuzzification process and also provided the forecasted

    values of better accuracy. The developed method is a generalized

    method of forecast thereby proving the forecast of better accuracy

    than the existing models.

    Satyendra Nath-Mandal [2008] has proposed the method to

    fuzzify the original data based on Gaussian function, triangular

    function, s-function, Trapezoidal and L –function. After fuzzifying, all

    fuzzified data were defuzzified to get normal form. The error analysis

    indicates that the membership function was appropriate for

    fuzzification of data and it was used to predict the short length at

    maturity.

  • 44

    Hui -Li Hsu et al., [2008] have presented evaluation techniques

    for interval forecasting which can provide a more objective decision

    space in interval forecasting to policy makers.

    Adesh Kumar Pandey et al., [2008] have proposed a comparative study

    of neural network and fuzzy time series forecasting techniques. It is

    successfully implemented for forecasting wheat production at pant

    Nagar farm.

    Chen et al., [2008] has proposed a comprehensive fuzzy time-

    series, which factors linear relationships between recent periods of

    stock prices and fuzzy logical relationships (nonlinear relationships)

    mined from time-series into forecasting processes.

    Dug Hun Hong [2005] has considered the expanded results to the non

    homogeneous fuzzy time series and the general fuzzy time series by

    using the weakest t-norm based algebraic fuzzy operations and solved

    the open problem.

    Taylor [2008] has used minute-by-minute British electricity

    demand observations to evaluate methods for prediction between 10

    and 30 minutes ahead.

    Muhammad Hisyam Lee et al., [2009] have proposed the

    adoption for the weighted and the difference between actual data

    toward midpoint interval based on fuzzy time series. The weights are

    determined according to chronological number of fuzzy logical

  • 45

    relations in fuzzy logic group and modification is also done in reversal

    of weight elements of transpose matrix for forecasting rule.

    Arkov et al., [2009] provided the method with high

    computational speed because it utilizes only operations of move and

    comparison. However, this approach allows the simulation of a limited

    number of systems with states depending on state quantization.

    Specific application is system state prediction.

    Nai-Yi-Wang et al., [2009] have presented a new method to

    predict the temperature and the Taiwan Futures Exchange (TAIFEX),

    based on automatic clustering techniques and two-factor high-order

    fuzzy time series. First, they had applied an automatic clustering

    algorithm to cluster the historical data into intervals of different

    lengths. Then, they applied the same based on two-factor high-order

    fuzzy time series.

    Kuo et al., [2009] have used the particle swarm optimization to

    find the proper content of the main factors. A new hybrid forecasting

    model which combined particle swarm optimization with fuzzy time

    series is proposed to improve the forecasted accuracy. The

    experimental results of forecasting enrollments of students of the

    University of Alabama found that the new model is better than any

    existing models, and it can get better quality solutions based on the

    first-order and the high-order fuzzy time series, respectively.

  • 46

    Chen et al., [2009] have presented a new method to forecast

    enrollments based on automatic clustering techniques and fuzzy

    logical relationships. First, he presented an automatic clustering

    algorithm for clustering the historical enrollments of the University of

    Alabama into intervals of different lengths. Then, each obtained

    interval is divided into p sub-intervals.

    Shiva Raj Singh [2009] has proposed a computational method of

    forecasting based on higher order fuzzy time series. The developed

    methods avoid the computations of complicated fuzzy logical relational

    matrices and search for suitable defuzzification method.

    Suresh et al., [2009] have presented a modified method of

    forecasting in fuzzy time series using transition probability

    membership function. In addition, the modified method is identified

    outlier in time series data using cook’s distance method.

    Okan Duru et al., [2010] have investigated the predictive

    performance of fuzzy time series analysis method for dry bulk freight

    method. The advantages of this method indicated the lack of several

    diagnostic tests like as normality and stationarity.

  • 47

    Narendra Kumar [2010] has developed time variant fuzzy time

    series models and its implementation testing for forecasting of wheat

    crop production. Also the results are compared with other known

    existing methods.

    Ismail et al., [2011] have discussed a modified weight for fuzzy

    time series and it shows a significant reduction of mean square error

    and average forecasting error when compared with known existing

    method.

    Lazim Abdullah et al., [2010] have presented a combination of

    fourth-order fuzzy time series with the multi-period adaptation model.

    It is adopted to forecast the KLCI stock index. The results found that

    the forecasting model that combines with the multi-period adaptation

    model produce lower RMSE compared existing approaches. It is

    proves that the multi-period adaptation models can effectively improve

    the forecasting performance.