This repository contains the code, data and visualization for the paper titled Modeling Intensification decision in Kilombero Valley Floodplain: A Bayesian Belief Network Approach. We developed a Bayesian Belief Network using Netica java API and application to model both the structure and parameters of the model. The structure is learned using TAN(Tree-augmented Naive Bayesian)¹ and the parameters are learned using Expectation maximization algorithm. To identify state specific factors we augmented the BBN approach with Design of experiment and regression tree. The repo also contains the data and a jupyter notebook for the regression tree analysis.

This study was conducted in the Kilombero Valley Floodplain (KVF), Tanzania. The low altitude plain with alluvial deposits has a productive natural resource base with fertile land, reliable water availability and extensive pastures (Bamford et al., 2010; Nindi et al., 2014). Located in the Ulanga and Kilombero Districts in southern Tanzania, it forms one of the four principal sub-basins of the Rufiji River Basin and comprises a myriad of rivers and seasonally flooded marshes and swamps (Dinesen, 2016).The seasonal change in water dynamic is huge and the plains sometimes becomes totally flooded during the wet season, while it dries up during the dry season with the exception of the rivers and river margins as well as of the areas with permanent swamps and water bodies (Kato, 2007; Ntongani, Munishi, More, & Kashaigili, 2014).

We conducted a household survey in 21 villages in two Districts of the Kilombero Valley, Ulanga and Kilombero. In total, 304 farm households were interviewed using a Standard questionnaire; giving their opinions upon a wide range of topics. The survey was designed to discover the farming system in terms of resources, land use and sources of livelihoods. In the links below you will find the data (for the paper), survey instrument and descriptive statistics:

• Data
• Questionnaire
• Descriptive statistics

We developed a bayesian belief network using Netica java API to model both the structure and parameters of the model. The structure is learned using TAN(Tree-agumented Naive Bayesian)algorithm and the parameters are learned using Expectation maximizationalgorithm. Netica provides a number of simplifying tasks for the modeler including a high visual capability to display the network and advanced algorithms to learn the structure and parameters of the network. Using the Java API for the construction of BBNs provides an advantage in terms of transparency, reproducibility and easy integration with other modeling tools of interest as it resulted in Java source code. The code is structures in to the following packages

• Build
• CrossValidation
• Display
• MainModel
• Sensitivity

To identify state specific factors we augmented the BBN approach with Design of expermint and regression tree. The repo also contains the data and a jupyter notebook for the regression tree analysis. Go Here for the code and detailed results of our sensistivity analysis.

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation,version 3 of the License. This program is distributed in the hope that it will be useful,but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You can find the copy of the GNU General Public License Here. The source code for NeticaJ is licensed under NORSYS LICENSE AGREEMENT FOR NETICA- Norsys Agreement #116. You can find the copy of the license Here

1: The structure is learned using the application version of Netica