PG Thesis
Permanent URI for this collectionhttp://localhost:4000/handle/123456789/2
Browse
18 results
Search Results
Item Design and analysis of best-worst scaling studies in agricultural research(Department of Agricultural Statistics, College of Agriculture, Vellayani, 2026) Anjana Bivas, TThe research study entitled “Design and analysis of Best-Worst Scaling studies in agricultural research” was undertaken at College of Agriculture, Vellayani, during 2023-2025. The primary objective of the study was to develop a comprehensive web application for designing suitable questionnaires and analysing data for Best-Worst Scaling (BWS) experiments in agriculture, guided by insights obtained from a bibliometric analysis of agricultural studies employing the BWS approach. In agriculture, effective decision-making depends on understanding the preferences and priorities of various stakeholders, including farmers, consumers, researchers, and policymakers. Traditional preference elicitation techniques often fall short in capturing subtle distinctions among choices. BWS offers a robust alternative for quantifying stakeholder preferences across domains such as technology adoption, agricultural policies, consumer behaviour, and resource prioritisation. A bibliometric analysis of agricultural BWS literature from 2011 to 2025 was conducted to identify how BWS has been applied, the experimental situations where the three cases of BWS are adopted, commonly used design methodologies, and the analytical approaches used. These findings provided clarity on current practices and guided what features and analytical capabilities needed to be prioritised in the system development for this research. Despite its growing use, researchers often face challenges in designing BWS questionnaires and analysing the resulting data, particularly when dealing with multiple attributes, complex profiles, or multi-profile choice sets. Manual generation of choice sets can be time-consuming and prone to errors, while advanced analytical models require considerable statistical expertise. These challenges underscore the need for an accessible and efficient tool to streamline both stages of BWS research. To address this gap, the web application, named PEAR-BWS (Preference Evaluation in Agricultural Research using Best-Worst Scaling), was developed using the R Shiny framework. It consists of two core modules, Questionnaire Generation and Statistical Analysis, offering an integrated environment for generating BWS questionnaires and analysing response data, without the need for programming skills. The Questionnaire Generation module enables users to build BWS-based choice sets for all three BWS cases (Object, Profile, and Multi-profile), ensuring balanced representation of items and profiles. The Statistical Analysis module integrates multiple analytical approaches, including Count Analysis, Multinomial Logit, Paired, Marginal, Marginal Sequential, Hierarchical Bayesian estimation, and Latent Class Analysis models. All questionnaire structures and analysis results generated from both modules can be downloaded as Word documents, facilitating direct use in research reporting, thesis writing, publication work, and field data collection. To demonstrate its analytical capabilities, three hypothetical model datasets were constructed for the three BWS cases, reflecting realistic response structures and consistent scoring (1 for best, -1 for worst, and 0 for others). An online survey conducted among students from different agricultural universities evaluated the usability and performance of the application. The feedback indicated a high level of user satisfaction, highlighting its efficiency and practical relevance. Overall, the study presents PEAR-BWS as a comprehensive and user-friendly tool that simplifies the design and analysis of BWS experiments, thereby enhancing accessibility and promoting evidence-based decision-making in agricultural research. The work provides a foundation that can be further expanded in the future by integrating more analytical methods with enhanced visualisation and direct data collection functionality within the web application.Item Web application for data visualization in agricultural research(Department of Agricultural Statistics, College of Agriculture, Vellayani, 2024-02-08) Burra Preeti; Pratheesh, P GopinathThe research project titled "Web Application for Data Visualization in Agricultural Research" was undertaken at the College of Agriculture, Vellayani, during the period from 2021 to 2023. The primary aim of this study was to develop a user-friendly web application for data visualization in agricultural research. In the realm of agricultural research and education, effective data visualization stands as a pivotal tool for comprehension and analysis. However, prevalent software tools in agricultural research such as R, Python, and Excel, while rich in capabilities, present formidable challenges hindering their widespread utilization in this domain. The complexities inherent in R code, runtime errors in Python, and the limitations of Excel often impede the seamless harnessing of their diverse visualization potential by agricultural students. In light of these challenges, there arose a significant demand for a specialized and user-friendly tool designed to meet the visualization needs of agricultural researchers. The findings from a pilot survey among agricultural students highlighted the challenges they faced with R's data visualization tools, despite its free availability and multiple data display options. Addressing the challenges faced by agricultural students using R's visualization tools, this research introduces grapesDraw, an open-source web tool and R package designed to democratize data visualization in agricultural research. Leveraging widely-used R packages such as ggplot2, dplyr, RColorBrewer, ggthemes, dplyr, packcircles, factoextra, grDevices, shinydashboard, shinyWidgets, shinycssloaders, shiny, algorithms for each plot were formulated, along with the development of corresponding User Interface (UI) and server modules. Individual testing of these modules was conducted, and upon successful verification, a basic app skeleton was constructed. Subsequently, the debugged components were transformed into a comprehensive dashboard. Following integration into a unified platform, thorough debugging processes were applied to ensure the stability and functionality of the app. By harnessing the robust capabilities of these R packages, grapesDraw not only addresses current visualization challenges but also lays a foundation for evolving with the dynamic needs of agricultural research visualization. Its modular architecture also facilitates easy expansion to incorporate additional tools for more comprehensive data visualization in the future. The design of this app skeleton adhered to a z-shaped pattern for web content viewing, reflects users' natural eye movement starting at the top-left, moving horizontally, then diagonally down to the bottom-left, and finally, scanning horizontally across the bottom. Adopting this layout strategy contributed to a more visually pleasing and user- friendly presentation of information within the web application interface. The grapesDraw is freely accessible through two different avenues: it operates as both a web application and an R package. While grapesDraw is hosted on shinyapps.io under the free tier plan, we recommend utilizing it primarily as the R package, which can be downloaded via github. Detailed instructions accompany the package, providing comprehensive guidance on its usage and functionalities. It plays a pivotal role not only in presenting research findings but also in facilitating effective communication through publications. A follow-up survey across various State Agricultural Universities assessed grapesDraw, where 75 per cent rated the web application as excellent, 23 per cent as good, and 2 per cent as fair. Criteria like user interaction and overall satisfaction were considered. Overall, 78 per cent rated the app as excellent, highlighting its strong approval among users. This research introduced grapesDraw, a user-friendly web application and R package, leveraging widely used packages like ggplot2 and shiny. Tailored algorithms within grapesDraw enhance precision for agricultural data visualization, complemented by a user-friendly z-shaped layout. Feedback from State Agricultural Universities highlights its success, with the majority rating it as excellent, emphasizing grapesDraw's crucial role in facilitating and enhancing agricultural data visualization.Item Artificial inelligent (AI) chatbot for scientific knowledge delivery in black pepper cultivation(Department of Agricultural Statistics, College of Agriculture, Vellayani, 2025-01-08) Blesson, B Varghese.; Pratheesh, P GopinathThe research work entitled “An artificial intelligent (AI) chatbot for scientific knowledge delivery in black pepper cultivation” was conducted at the College of Agriculture, Vellayani, Thiruvananthapuram, Kerala, from 2022 to 2024. The objectives of the study were to develop an intelligent chatbot specifically tailored to the needs of farmers and extension workers in black pepper cultivation and to assess the suitability of an embedding model and a fine-tuned model of OpenAI for developing the chatbot. A detailed survey was conducted in Idukki and Wayanad, selecting 60 farmers with the assistance of the Department of Agriculture Development and Farmers’ Welfare, Government of Kerala. The survey focused on gathering critical data about the key concerns of black pepper farmers enabling us to build a comprehensive chatbot database. The survey outcomes highlighted major challenges faced by the farmers are disease management, pest management, lack of awareness about government schemes, insufficient extension services, nutrient management and climate change-related issues. These critical findings laid the foundation for the subsequent development of a comprehensive chatbot database tailored to address these farmers identified issues effectively. The database was then enriched by incorporating black pepper cultivation practices recommended by Kerala Agricultural University (KAU) alongside crop management strategies from the Indian Institute of Spices Research (IISR) as standard references. This approach combined direct farmer perspectives with scientifically backed methods creating a robust database that the chatbot could use to deliver accurate responses to farmer queries. The process of data preparation included key preprocessing steps to improve model performance. Techniques such as text cleaning, lowercasing, stop word removal and stemming were done sequentially. These steps were crucial for reducing data noise, simplifying complexity, minimizing overfitting and ensuring that the chatbot could generalize across varied questions. The development of the chatbot started with generating the necessary token keys for Telegram and OpenAI. The embedded model chatbot is designed to answer questions related to black pepper cultivation by leveraging the text-embedding-ada-002 model. The embedding model converts the query of the user into embeddings and matches the query to the database to retrieve relevant answers. A custom fine-tuned chatbot was developed using OpenAI’s GPT-3.5 turbo-0125 model. Both models were integrated into Telegram as Telegram is free, cross-compatible and has unlimited user limits. The study employed a quantitative research methodology to assess chatbot acceptance and compare models using a survey based on the Use and Gratification Model (UGM). The survey included 150 participants, comprising farmers, scientists, agriculture officers and research scholars. The performance of each model was assessed across three dimensions: technology, hedonics and risk. Technology was evaluated for authenticity and convenience, hedonic for enjoyment and entertainment and risk for privacy concern and immature technology. Behavioural intention was the dependent variable. The demographic analysis showed that most respondents were aged between 28 and 35. The reliability of the survey scale was confirmed through Cronbach's alpha. The Mann-Whitney U test and regression analysis, provided insights into the comparative performance of the two models. The Mann-Whitney U test highlighted significant differences in three areas: authenticity, immature technology and behavioural intention. The embedding model had a higher mean score for authenticity and behavioural intention, while the fine-tuned model exhibited higher concerns with immature technology. The multiple linear regression analysis was done for both models keeping behaviour intention as the dependent variable. Under both regression models, authenticity of conversation, privacy concern and immature technology were found to be significant. Authenticity of conversation has a slightly higher impact in the embedding model compared to the fine-tuned model. Privacy concern has a negative relationship in both models. Immature technology shows a stronger negative effect in the fine-tuned model compared to the embedding model. Based on the statistical analysis the embedding model was found to be better than fine-tuned model. The embedding-based model's adaptability and broad language representation made it ideal for diverse queries, effectively addressing farmer concerns and enhancing user satisfaction. The model’s strength in similarity based search provided a robust solution proving more versatile than the fine-tuned model for real-time, varied interactions in agriculture.Item Forcasting major pests of rice using weather variables(Department of Agricultural Statistics, College of Agriculture,Vellanikkara, 2025-03-06) Abishek Krishnan.Rice is the staple food crop of majority of the global population and production of rice has been affected by various factors. Insect pest attack in rice causes significant yield loss in rice. Forecasting models can be developed to determine the incidence of the pest using various weather parameters and it is a significant step in helping the farmers to mitigate the loss of crop due to insects. The forecasting of two major insect pests of rice namely Yellow stem borer (YSB) and Brown plant hopper (BPH) has been taken up in this study. Population count data of YSB and BPH from light trap catches installed at Regional Agricultural Research Station (RARS), Pattambi, Kerala was utilised for the study. The objectives of the study were to establish the relationship of YSB and BPH population with the weather variables which includes maximum temperature (TMAX), minimum temperature (TMIN), morning relative humidity (RH I), evening relative humidity (RH II), rainfall (RF) and sunshine hours (SSH); to develop suitable forecasting models for YSB and BPH using weather variables; and to study the dynamics of these pests under pre and post flood conditions. Two period of peak incidence of the YSB population was revealed from the analysis of their distribution during the study period ranging from 1997 to 2023. These periods of peak incidence were 10th to 16th standard meteorological week (SMW) and 38th to 48th SMW. Relationship of pest population with weather variables was studied using correlation analysis. The peak week corresponding to the 12th SMW had significant negative correlation with TMAX, TMIN, RH I, RH II and SSH, whereas count of YSB for the 43rd SMW had an association with TMIN, RH I and RH II. The weekly distribution of BPH showed one prominent peak incidence corresponding to the 39th to 44th SMW and it was found that minimum temperature, morning and evening relative humidity, rainfall, sunshine hours were the weather variables that had significant correlation with the BPH population. Multiple linear regression model for to predict the incidence of YSB at 12th week of peak incidence, had weather variables TMAX, TMIN, RH I and SSH as explanatory variables and the model yielded an adjusted R2 of 81%. The interaction of temperature and relative humidity was found to be significant in case of composite regression models. Multinomial logistic regression models developed for determining the pest infestation status, in terms of low, medium and high incidence of the YSB population had an accuracy of 92% and 72% for the 12th and 43rd peak weeks respectively. In case of BPH population, multiple linear regression model to predict BPH incidence during 39th – 44th SMW yielded a model with adjusted R2 value of 39% with weather variables TMIN, RH I and RH II. For 42nd SMW, model with variables TMIN, RH I and RH II yielded an R2 value of 45%. In case of BPH also, interaction of temperature and relative humidity was found to be significant in the fitted composite regression models, with a greater model R2 in comparison to that of multiple linear regression model. An accuracy of 66% and 64% was obtained for the model fitted using multinomial logistic regression analysis during peak period and peak week respectively. Autoregressive Integrated Moving Average (ARIMA) model, ARIMA with exogenous variables (ARIMAX) and Integer valued generalized autoregressive conditional heteroscedasticity with exogenous variables (INGARCHX) models were employed for the time series analysis of the pest data. Three different trend periods were identified for both YSB and BPH population. In the case of YSB population during first period, ARIMAX (2,1,1) was the best fit model with MAPE value of 11.29%. For second period, ARIMAX (2,1,1) with MAPE of 11.36% and for the third period, ARIMAX (2,0,1) with a MAPE of 15.02% were the best fitted models. In the case of BPH, ARIMAX (2,0,2) was the best fit model for first period, ARIMAX (2,1,2) and ARIMAX (1,1,1) for the second and third period respectively. INGARCHX models fitted for YSB and BPH population were of poor fit, as these models did not have any significant weather variables in the model, with only model parameters turning out to be significant. Further, pre and post flood analysis of pest dynamics with weather variables revealed that there is a shift in the peak period of incidence for both YSB and BPH and further there is an overall decline in both YSB and BPH population for post flood period. The individual and joint effect of the weather variables were determined using the multiple linear regression analysis and composite regression analysis respectively, wherein composite regression provided better model accuracy in comparison to multiple linear regression analysis. Multinomial logistic regression analysis helped to determine the epidemic status of YSB and BPH at different peak periods of incidence. These epidemic status can be used as a warning alert for the farmers on the level of incidence of these pests. Time series analysis of the pest population count revealed that ARIMAX models performed better than INGARCHX models in predicting population of YSB and BPH and it also shed light into the various trend breaks observed in the pest population count for both YSB and BPH across the time period studied.Item Development of software for statistical methods in social science research(Department of Agricultural Statistics, College of Agriculture, Vellayani, 2022-01-12) Sandra, M M.; Brigit JosephThe research entitled “Development of software for statistical methods in social science research” was conducted during the period 2020-22 in KAU, COA Vellayani. The objective of the study was to develop an open-source software for social science research with special focus on survey data analysis in agriculture. The methods covered included canonical correlation analysis (CCA), linear regression, binary logistic regression, chi-square test, index construction, test for scale reliability, and construction of one-way frequency tables for quantitative data. Scales are measurement tools used by social scientists to measure phenomena of abstract nature. They are collections of questions, the responses for which are used to measure the construct under consideration. A developed scale should be tested for its reliability before it can be put to use. To facilitate this, an application was developed for testing the temporal stability (test-retest reliability) and internal consistency (Spearman-Brown prophecy formula and Cronbach’s alpha) which returns the test statistics and their significance. Indices are constructed so as to condense complex multidimensional phenomena into a simpler form to make evaluation easier. Indicators are unidimensional data extracted from the sample. They are combined using various method to form indices. An application was developed for the construction of indices after standardizing the dimensions and aggregating with equal weights. Frequency tables are very popular analysis tools for data collected from a sample survey. In one-way tables, respondents are classified into different categories based on different levels of a single factor. The percentages are also calculated for describing the nature of the sample. An application was developed that calculates and displays the frequencies and percentages based on a prescribed criterion for classification of quantitative data. Chi-square statistics can be used for analyzing categorical data. The goodness of fit test employs the chi-square statistic to conclude if the observed frequencies are on par with theoretical frequencies whereas test for independence of attributes is employed to check whether two attributes are distributed independently of each other in a sample. An application was for conducting the two tests mentioned. After 140 140 uploading the data in the prescribed format, the test statistic along with p-value is returned. Regression involves studying the functional relationship between a single dependent variable and one or more independent variables. Two types of regression is considered based on the nature of the dependent variable, whether they are quantitative or qualitative. Separate applications are developed for both types. The regression model developed in either case is assessed based on the value of coefficients. In the case of linear regression, the dependent variable is qualitative and the model is linear with respect to parameters. Ordinary least square method is used to estimate the parameters of the model and their significance is tested using t-test. The goodness of fit is assessed using R-squared and adjusted R-squared values which represent the proportion of the variance in y explained by x variables. When the dependent variable is qualitative it is called logistic regression. When the regressand has only two possibilities it is called binary logistic regression. In this case also the coefficients of each independent variable and their significance in terms of p-values are provided. Canonical correlation analysis (CCA) is a dimension reduction technique in which the relationship between two multivariate datasets can be summarized into a few significant dimensions. The datasets are linearly combined to obtain pairs of canonical variables. An application was developed for the CCA. The results include the correlation between the pairs of canonical variates called canonical correlations. There are as many correlations as there are number of variables in the smaller group. The significance of dimensions/ correlations can be assessed using Wilks’ lambda criteria and associated p-value. The application also returns raw and standardized coefficients of canonical variates in each dimension and also loadings and cross loadings. The open-source language R and its associated integrated development environment RStudio was used along with the web development platform RShiny was used for the development of the application. The results were demonstrated using example datasets. The software covers statistical methods that is frequently used in researches in social sciences.Item Comparison of time series and machine learning (ML) models for pinapple price forecasting in Kerala(Department of Agricultural Statistics, College of Agriculture, Vellayani, 2025-01-28) Arshida, A K; Manju Mary PaulThe research work entitled “Comparison of time series and Machine Learning (ML) models for pineapple price forecasting in Kerala” was conducted at the College of Agriculture, Vellayani, Thiruvananthapuram, Kerala, from 2022 to 2024. The objectives of the study were to develop time series and Machine Learning (ML) models for price forecasting and comparison to determine the best performing model in pineapple price forecasting. Pineapple (Ananas comosus, Bromileacea) is one of the commercially important fruit crops of India, driven by it's popularity, nutritional benefits, and adaptability. Daily price data of both green and ripe pineapple for the past 18 years starting from January 2006 to December 2023 were collected from the website of Vazhakulam Pineapple Growers Association (VGA), Kerala. A comprehensive preliminary analysis was done in the collected data, which revealed significant autocorrelation, seasonality, and volatility in the same. In order to understand the behaviour of the time series, various statistical tests, such as those for normality and stationarity, were conducted. The data was pre-processed to handle missing values before model development. The dataset was then divided into training and test sets, with 80 percent of the data allocated for training the models and 20 percent reserved for testing their predictive performance. Several forecasting models were developed which include traditional time series models such as Autoregressive Integrated Moving Average (ARIMA), Generalized Auto-Regressive Conditional Heteroscedasticity (GARCH) and machine learning models such as General Regression Neural Network (GRNN), Support Vector Regression (SVR), Long Short-Term Memory (LSTM) networks and Facebook Prophet. Parameter tuning was done to optimize the hyperparameters of a model to improve its performance and to achieve the best possible accuracy for each of the models that were developed. The performance of these models was evaluated using multiple error metrics, including Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). These metrics provided a comprehensive view of the models' accuracy and robustness in forecasting pineapple prices. The results revealed that the LSTM model consistently outperformed all other models in terms of predictive accuracy, exhibiting the lowest error rates across all performance metrics. This superior performance can be attributed to LSTM’s ability to model long-term dependencies and capture intricate temporal patterns in the price data. Models like SVR and GRNN also showed better performance when compared to ARIMA and GARCH. On the other hand, the ARIMA model, while useful for short term predictions, displayed lower accuracy compared to other models, likely due to its limitations in handling non-linearities and complex seasonal patterns. The comparative analysis of the different models shows the effectiveness of machine learning techniques over traditional time series methods in price forecasting.Item Rainfall probability analysis for Kerala(Department of Agricultural Statistics, College of Horticulture, Vellanikkara, 2020-08-11) Pooja, A; Laly John, CThe study entitled “Rainfall Probability Analysis for Kerala”, was conducted to estimate the probabilities and amounts of normal, excess and deficit rainfall during southwest and northeast monsoon periods, to determine the conditional probabilities for monsoon to be normal, excess and deficit when the monthly rainfall is normal, excess and deficit and to determine the assured monthly and weekly rainfall in the five agroclimatic zones of Kerala. Daily rainfall data from 1983 to 2019 collected from the five stations viz. Vellayani, Kumarakom, Vellanikkara, Pilicode and Ambalavayal representing southern zone, problem area zone, central zone, northern zone and high range zone respectively formed the database for the study. The daily rainfall data were converted to annual, seasonal, monthly and weekly rainfall data series for further analysis. Significant increasing trend is observed in annual rainfall at Vellayani. Non-significant increasing trend in annual rainfall is also observed at Vellanikkara, Pilicode and Ambalavayal. But, Kumarakom exhibited a non-significant negative trend for annual rainfall. Except the stations of Kumarakom and Vellanikkara, other stations have non-significant increasing trend in southwest monsoon rainfall. Only Vellayani has a non-significant increasing trend in northeast monsoon rainfall. All other stations exhibit declining trend for northeast monsoon rainfall over the period 1983-2019. The exploratory analysis showed that wide variation exists in rainfall over different agroclimatic zones. The rainfall variability was visually made possible by the use of box-and-whisker plots constructed for annual, seasonal, monthly and weekly rainfall over five stations. The boxplots explained the average and dispersion of rainfall data in terms of quartiles. In all the stations, annual, southwest monsoon and northeast monsoon rainfall did not follow normal distribution which was evident from the uneven boxes and dissimilar whisker lengths. A station-wise comparison showed that annual rainfall as well as southwest monsoon rainfall increased from south (Vellayani) to north (Pilicode) of Kerala and northeast monsoon rainfall declined from south to north. Outliers representing extreme rainfall events were also detected using boxplots which made the rainfall distribution skewed. At Ambalavayal, extreme annual rainfall (3093 mm) and extreme southwest monsoon rainfall (2380 mm) happened in 2018 and at Pilicode, extreme annual rainfall (4803 mm) occurred in 1994.The northeast monsoon witnessed extreme rainfall in 1992 (764 mm) and 2010 (950 mm), at Vellanikkara. The analysis of monthly rainfall using boxplots clearly depicted bimodal distribution at Vellayani and unimodal distribution at Pilicode. Thus, it was appropriate to use median and Inter Quartile Range (IQR) as the measures of central tendency and dispersion, rather than mean and standard deviation, as former measures were not affected by extreme observations. Conditional probabilities were estimated to determine whether the seasonal rainfall was excess, normal or deficient when monthly rainfall was excess, normal and deficient. Excess, normal and deficient rainfalls were defined in terms of quartiles for the purpose. At Vellayani, there was more than 50 per cent probability for the southwest monsoon to be in excess if the rainfall in the month of June was excess. At Kumarakom, there was a 60 per cent chance for receiving excess rainfall in southwest monsoon when July and September had excess rainfall. Rainfall in June and July determined excess rainfall in southwest monsoon at Vellanikkara with 70 per cent probability. There was 70 per cent chance to receive excess rainfall during southwest monsoon at Pilicode when August or September received excess rainfall. Ambalavayal had experienced excess rainfall with 80 per cent possibility when July had excess rainfall. Similarly, each station had varying effects of rainfall in October and November towards northeast monsoon season. This analysis can provide information about when to take precautionary measures to anticipate the threat of floods and droughts in near future. The assured amounts of rainfall varied during different months and weeks in each station. Selection of crops (short duration varieties for low rainfall regions), cropping pattern and other decision making strategies for harvesting rainfall or drainage management can be made on the basis of assured rainfall in different months. At Vellayani, 23rd (4th June to 10th June) and 24th (11th June to 17th June) standard meteorological weeks (SMWs) are the rainiest weeks. The rainiest week is 28th SMW (9th July to 15th July) at Kumarakom and Ambalavayal, whereas, it is 24th SMW (11th June to 17th June) at Vellanikkara and Pilicode. On farm operations and crop management plans like decision of times of sowing, irrigation, fertilizer application, harvesting, etc., could be executed on the basis of assured weekly rainfall during the different crop periods.Item Comparative analysis of price forecasting models for black pepper(Department of Agricultural Statistics, College of Agriculture,Vellanikkara, 2024-02-17) Akshaya Ajith.; Sajitha Vijayan, M; Srinivasan, KBlack pepper, the ‘king of spices’ is one of the most popular and widely consumed spices which shares a place on most dinner tables with salt. India is the third largest producer in the world (International Pepper Community, 2023), also a significant consumer and exporter of black pepper, with Kerala and Karnataka producing the majority of the nation's output. Kerala ranked second in terms of black pepper acreage (76,160 ha), and production (33290 metric tonnes) but seventh in terms of productivity (0.44 metric tonnes/ha) (GOI, 2023). Historically, the market value of pepper contributed to the development of the city of Kochi as a centre of international commerce. Kochi has the first exclusive pepper exchange in India which was established by the Indian Pepper and Spice Traders Association, IPSTA and the exchange was well regulated by the traditional players here, without any default on supply or delivery of the commodity and volatility. As per latest records of Spices Board, the price of black pepper surpassed Rs.500/Kg which was the maximum price in the past years since 2022 in Kochi market and this indicated that black pepper prices are highly volatile. Being a perennial crop, the large variation in prices of black pepper within a year is a major problem faced by farmers as well as consumers. Hence, analysis of time series data of prices of black pepper is of prime importance. In this context, the present study was undertaken to evaluate different time series models for prices of black pepper and to suggest suitable forecast models for Kochi market. Time series data on monthly and weekly average prices of garbled and ungarbled black pepper at Kochi market from January 2000 to December 2020, collected from Spices Board, Kochi formed the database for the study. Analysis of price pattern revealed that wide fluctuation exists in the prices of black pepper in Kochi market. In order to have a general idea about trend of prices of black pepper, models like linear, exponential and quadratic, were fitted. From among several models tried, exponential model was found to be best fit for the monthly and weekly prices of both garbled and ungarbled black pepper. To examine the time series data for the price of black pepper, a multiplicative model was employed for decomposition. Seasonal indices for the 12 months from January to December was calculated for both garbled and ungarbled black pepper prices as the seasonal variation were present in monthly and weekly data. Different traditional time series models such as exponential smoothing models (single, double, Holt-Winters’ multiplicative models (HWMS)), Auto Regressive Integrated Moving Average (ARIMA), Seasonal ARIMA (SARIMA), Auto Regressive Conditional Heteroskedasticity (ARCH) and Generalized Auto Regressive Conditional Heteroskedasticity (GARCH) were applied to the price data using R software. The Augmented Dickey-Fuller test and Heteroscedasticity Lagrange’s Multiplier test were used to test the stationarity and volatility of the time-series respectively. In addition to the traditional method of price forecasting, machine learning techniques like Artificial Neural Network (ANN), and Recurrent Neural Network (RNN) were applied to the data. An artificial neural network (ANN) is a mathematical model that aims to replicate the functionality and architecture of biological neural networks. Time- delay Neural Network (TDNN) which is a major type of ANN was employed for this temporal data as it uses time delays at the input layer of the network to build a short-term memory model for forecasting the prices of black pepper. An artificial neural network with a recurrent topology is called a recurrent neural network. Long Short-Term Memory (LSTM) neural network, a specialized type of RNN, as it possesses the capability to learn patterns with long dependencies and is adept at detecting complex patterns. The price data was split into training and testing data in the ratio of 80:20. The best forecasting model was determined based on the lower values of the Root Mean Square (RMSE) and Mean Absolute Percentage Error (MAPE). The predictability performance of the selected model was also evaluated using MAPE. The HWMS model was observed as the finest among the different exponential smoothing models for this time series data on prices of garbled and ungarbled black pepper. SARIMA(2,1,2)(3,0,2)12 , SARIMA(2,1,2)(2,0,2) 12, SARIMA(2,1,2) (1,0,0)52 and SARIMA(1,1,1) (1,0,1)52 were identified best among several ARIMA models tried for monthly and weekly garbled and ungarbled black pepper respectively. The GARCH (1,1) was considered best among the different ARCH family models for this price series data. TDNN (6:2s:1l), TDNN(6:3s:1l), TDNN (13:8s:1l) and TDNN (12:7s:1l) models were found to be the pinnacle artificial neural network model with lower MAPE values 4.29, 4.63, 1.99 and 2.22 in the case of monthly and weekly prices of garbled and ungarbled black pepper respectively. The results revealed that the TDNN model showed superiority in price forecasting of black pepper in all the cases when compared with all other models. Thus, the TDNN model was used to forecast the prices from January 2021 to December 2022. The MAPE value between the actual and forecasted prices for 2021 and 2022 was 4.19 and 4.86 respectively for monthly price of garbled black pepper, while for monthly price of ungarbled black pepper it was 4.09 and 5.05 respectively. The MAPE value between the actual and forecasted prices for 2021 and 2022 was 3.11 and 3.16 respectively for weekly price of garbled black pepper, while for weekly price of ungarbled black pepper it was 3.36 and 3.58 respectively. In conclusion, the analysis suggested that the TDNN model proves to be a reliable forecasting tool for predicting prices of black pepper in the Kochi market. The robustness of the TDNN model offers a plethora of opportunities for understanding the future price pattern of black pepper which enables, various stakeholders such as producers and traders to adapt with the price fluctuations and for policymakers to ensure market stability. The TDNN model's reliability in forecasting black pepper prices not only enhances market transparency fostering overall market efficiency in India.Item Statistical assessment of banana ripening using smartphone - based images(Department of Agricultural Statistics, College of Agriculture ,Vellayani, 2022) Haritha R Nair; Pratheesh P GopinathItem Geostatistical analysis of groundwater level in Thiruvananthapuram District(Department of Agricultural Statistics, College of Agiculture, Vellayani, 2022) Harinath, A; Pratheesh P GopinathThe research work entitled “Geostatistical analysis of groundwater level in Thiruvananthapuram district” was carried out at the College of Agriculture, Vellayani during 2019-2021. The objective of the study was to analyze the spatiotemporal variations in the groundwater level, identify the relationship between groundwater and climatic factors (i.e., rainfall and temperature), and to prepare the thematic map for the location. To characterize the spatiotemporal fluctuations in groundwater level within the research region, various geostatistical approaches were used. The WRIS [Water Resource Information System] website was used to collect groundwater level data for 29 different locations within the study area for 10 years, from 2008 to 2017. The selection of data points was based on the even spatial distribution such that all the locations in the district are entirely covered. The NASA satellite website data was used to collect the rainfall and temperature data for the 29 distinct sites throughout a ten-year period. The semivariogram models were fitted to assess the spatial continuity of groundwater level. The nugget to sill ratio is also identified for detecting the spatial dependency. In the research region, the kriging interpolation approach was used to assess the spatiotemporal fluctuations in groundwater levels. If the data sets are normally distributed, the kriging interpolation technique will be more successful. Thus, the data points were subjected to exploratory data analysis to test the normality of the data set. The normality of the data sets is found out by Shapiro-Wilk’s normality test. The results showed that the years 2010 and 2017 are not normally distributed as the null hypothesis of the test is rejected. And also, in the case of temperature and rainfall, all the data points were not normally distributed. Thus, for the proper analysis, the log transformation was performed to the data sets which are not normally distributed and proceeded to further steps. The relationship of groundwater and climatic factors were accounted with the correlation analysis. The results showed that the temperature is having more dependency with the groundwater level fluctuation than the rainfall. 88 The semivariogram fitting were done to the groundwater level drop for each location, groundwater level over the years, and for the average groundwater level to identify the spatial and temporal variations in the study area. The drop was found out for each location by taking the difference between the groundwater levels of the years 2008 and 2017. The positive drop refers the depletion in the groundwater level and the negative drop refers the increment in the groundwater level. The nugget to sill ratio explains that the groundwater level drop is having a relatively strong spatial dependence. The three models, Spherical, Exponential and Gaussian models were fitted to the groundwater level for each year. The best fit model was selected by accounting the Adjusted R2 value. The spatiotemporal variation was studied by kriging interpolation method. The thematic maps were created to analyze the groundwater level variations. The maps were created in the ArcGIS 10.4 software. By investigating the maps prepared, the groundwater level depletion is observed severely in the Varkala region, and the Parassala region. The groundwater level at the high ranges like Ponmudi, Bonacaud, and Neyyar regions are maintaining a decent amount of groundwater level. From the PCA biplots prepared, the study concluded that there is a gradual groundwater depletion happening from 2008 to 2017. And from the biplot of years, the temperature is relatively high in 2016, 2017 where the groundwater level is also high. And the temperature is relatively low in 2008, 2009 where the groundwater level is also low. Thus, it can be concluded that the groundwater is having some dependency with the temperature variations which have been detected in the correlation analysis. From the biplot of different locations, it can be analyzed that the Varkala, Sreekariyam, Pothencode, Chengal, Neyyattinkara regions are having high groundwater depth. And Kattakkada, Kallar, Palode, Ariyanadu, Maruthamoola, Peringamala regions are having low groundwater depth. From the research performed, it can be concluded that, most of the locations are having a positive drop in the groundwater, which represents that the groundwater depletion is happening in temporal structure in the study area. The highest depletion in the 89 groundwater is seen in Pothencode, Chengal, Varkala, Neyyattinkara regions. The rate of groundwater level drop is 1.49 meters, which is positive, and can be inferred that there is depletion in the groundwater level. The nugget to sill ratio of the groundwater level drop in the study area is 0.367, which refers that the depletion is moderately spatially dependent. From the correlation analysis, it can be concluded that the temperature is a major factor influencing the groundwater depletion than the rainfall, because there is a positive significant correlation between groundwater and temperature. The groundwater depth of Varkala, Pothencode, Sreekariyam, Neyyattinkara, Chenkal, Kulathoor is high, and at Kattakkada, Palode, Kallar, Ariyanadu have low groundwater depth which can be concluded from PCA biplot of different locations