Efficient and reliable forecasting techniques for commodities with volatile price series are indispensable in agriculture dependent country like India. In this context, choosing a universally accepted model for forecasting precisely the price series of commodities like onion is one of the most challenging tasks because of the existence of seasonality, non-linearity and complexity in the data, simultaneously. Time series models like GARCH, machine learning techniques like TDNN, SVM and deep learning models like LSTM, Stacked LSTM and Bi-LSTM have been extensively studied in this research work to judge their performance on volatile weekly price series of onion for two different markets in India. The models were tuned with the training dataset to forecast the values for the next twelve horizons and eventually the forecasted values have been compared with the testing dataset. It was found that deep learning

  In survey sampling, auxiliary information is often used to increase the efficiency of estimators of finite population parameters. The Calibration Approach is one of the popular techniques for such purposes. In this current study, product type calibration estimator of the finite population total has been proposed following well-known Calibration Approach for the situations of availability of inversely related auxiliary information at PSU level under two stage sampling design framework. Statistical properties of proposed product type calibration estimator of population total were studied through a simulation study. The simulation results suggest that the proposed product type calibration estimator is performing better than usual Horvitz-Thompson and linear regression estimators of the population total under two stage sampling design.

  Population ratio is one of the most commonly used statistics in official statistics, agriculture and agricultural-related fields. When interest is in the estimation of ratio for rare but highly aggregated geographically distributed population, adaptive cluster sampling (ACS) design is usually used (Dryver and Chang 2007). Under ACS design, neighbouring units are added to the sample if it satisfies a pre-determined criterion. ACS design allow observed values to trigger increased sampling effort during the survey. This intuitively appealing design can have lower variance than conventional designs. In many sampling survey situations, certain auxiliary information is often available and used for increasing the precision of estimator. Calibration approach given by Deville and S?rndal (1992) is widely used technique for this purpose. In this article, calibration estimator of population ratio under adaptive sampling has been developed when auxiliary variables are known. The variance and the estimate of variance for these estimators are obtained. The statistical performance of the proposed calibration estimators of population ratio under ACS were evaluated through a simulation study based on real population data with respect to conventional Horvitz Thomson (HT) estimator of population ratio which do not utilize the auxiliary information. The results of the simulation study show that proposed calibration estimators are more efficient than conventional HT estimator of the population mean under ACS with respect to percentage Relative Root Mean Squared Error (%RRMSE).

  In an agriculturally depending country like India, accurate and reliable drought forecasting is very important to figure out how drought will affect water resources and agriculture. Data-driven machine learning forecasting techniques are promising approaches for drought forecasting since they take less development time, fewer inputs, and are less sophisticated than dynamic or physical models. Machine learning models for drought forecasting use drought indices that are more operational than raw climatic variables. In this study, the potential of wavelet-based extreme learning machine (W-ELM) model to forecast effective drought index has been explored for Sagar and Chattarpur districts of the Bundelkhand region of India.The performance of W-ELM model has been compared with the other competitive machine learning models like support vector machine (SVM), extreme learning machine (ELM), and artificial neural network (ANN). Observational outcomes reveals that the W-ELM model outperforms ELM, SVM, and ANN.

  The economy of the state Kerala is dominated by agriculture and the agriculture depends on rainfall. Hence, the study of rainfall is important and its forecasting will aid in crop and hydrological planning. The present study analyzed the pattern of rainfall in Kerala for the period 1991 to 2020 and obtained annual average rainfall as 2906.79 mm. There was no significant trend in annual rainfall. As per monthly rainfall data, month of June receives highest rainfall followed by July. Monthly rainfall had been modelled using Seasonal Autoregressive Moving Average model (SARIMA) and Neural Network Auto Regression (NNAR) model. Comparison of models based on the accuracy measures, revealed NNAR (6,1,4)[12] as the best model for forecasting rainfall in Kerala. Monthly rainfall for 2021 and 2022 was predicted and it showed that rainfall will be high in the month of July.

  Time series modelling utilizes previous values to forecast the future values. Exponential smoothing is one of the approaches to make forecast as well as to smooth the time series data. Among the various exponential smoothing model, Single Exponential Smoothing (SES) is the most popular model in time series due its simplicity of understanding and implementation. On the other hand, state space methodology is a very useful technique to solve various problems in time series which is required to improve a system over time. This state space methodology can be used to represent various time series models including Autoregressive Integrated Moving Average (ARIMA). Kalman filter technique is an approach to estimate the time-dependent parameters. One heartening feature of Kalman filter is that it provides the minimum mean squared error (MSE) estimates for linear model. In present study, an attempt has been made to represent the SES in state space form and parameters are estimated using Kalman filter in conjunction with prediction error decomposition form of the likelihood function. An illustration has been given with different applications in agricultural domain. It has been seen that state space form of SES provides lower MSE compared to traditional SES. This integration of SES with state space formulations in agricultural domain will open a new era in agricultural modelling and forecasting.

  A four-associate class association scheme named as Dichotomized Split-Set (DiSS) association scheme is defined for v = 2(p-1)p number of treatments and a method for constructing Partially Balanced Incomplete Block (PBIB) designs based on this association scheme is developed. The proposed designs are cost effective in terms of resources as they require lesser replications. They are resolvable; hence they possess high application potential in areas like multi-site varietal trials where experimenters generally prefer incomplete block designs. The efficiency factors for these designs are computed in comparison to an orthogonal block design and are found to be quite high. For easy generation of these designs for any v = 2p(p - 1); p ≥ 3, an R package called ?ResPBIBD? is developed.

  conditions.The study examined 35 years of time series data on wheat yield as well as weekly data on five weather variables for the crop season from 1985-86 to 2019-20. Using weather indices and time trend as regressor variables and wheat yield as regressand the effect of various factors was investigated using step-wise regression analysis. It has been found that weighted weather indices of each weather variable including time trend have exhibited significant effect on the wheat yield. It has also been found that rise in all five weather variables except relative humidity has been detrimental to wheat yield during harvesting phase of the crop.The overall results indicate the fact that changes in climatic variables show detrimental as well as beneficial the role depending upon the phases of crop production in getting out its final output. On the basis of root mean square error the Model-P5 has been proven to be best among all the models the average percent standard error (PSE) value of the Model-P5 is 0.94 which shows that these models are better for forecast.principal component techniques are best created model.

  The ratio in finite population is one of the most common statistics used in official statistics, demographic studies, agriculture and allied field of agriculture. In this paper, estimators of the ratio/proportion in finite population are developed by incorporating known auxiliary information under the calibration approach. The variance and the estimate of variance for these estimators are obtained. A simulation study is carried out to evaluate the performance of proposed estimators comparing them with a simple estimator of the Population ratio that does not incorporate auxiliary information.

  Row-column designs are useful for the experimental situations in which there are two cross classified sources of heterogeneity in the experimental material. Often it is desired to compare two or more factors in row-column set up where only two units can be accommodated in a single column. In this article, a general method of construction has been developed to generate row-column designs for factorial experiments with two rows which permit orthogonal estimation of all main effects and specific two factor interactions as per the choice of experimenters.

  NA

  NA