Search

Research in educational data mining to establish or predict the retention of students in higher education institutions, as well as predict graduation performance abounds. This research is a data mining based project aimed at generating a model that can be used for predicting student's ability to graduate on time. In this research we have examined various factors such as age, gender, continuous assessment results, and final exam results, determine how they influence a student's graduation schedule. We have demonstrated our application of classification as a data mining technique to identify interesting patterns, and subsequently use predictive techniques to predict the possible consequent outcome, and further have conducted a detailed examination of the J48, Bayes Net, PART and Random Forest predictive algorithms and compared to draw conclusions on the data mining prediction tools that give optimum results. The J48 stood out in terms of performance output.

There is a huge generation of raw data during production processes of refinery products and in most cases this data remains under-utilised for knowledge acquisition and decision making. The purpose of this study was to demonstrate how data mining techniques can be used to develop models to predict product quality properties for petroleum products. This study used petroleum refinery production raw data to build predicting models for product quality control activities. The plant and laboratory data for the period of about 18 months was mined from the refinery repositories in order to build the datasets required for analysis using Orange3 data mining software. Four data mining algorithms were chosen for experiments in order to determine the best predicting model using cross-validation technique as a validation method. This study only employed two measuring metrics, classification accuracy (CA) and root mean square error (RMSE) as performance indicators. Random forest came out as the best performing model suitable for predicting both categorical (CA) and numeric data (RMSE). The study was also able to establish the relationship between the variables that could be used in critical operational decisions.

Mobile money has been known to be a successful venture around the world especially so, for African countries due to the many limitations that traditional banks have like operations, expensive transaction costs and cumbersome process to open account to mention but a few. The presence of mobile money has not only allowed the unbanked population to have accounts but has also alleviated poverty for many rural communities. Zambia has seen an increase of mobile money accounts and COVID-19 has exacerbated this increase. Therefore, this paper sought to determine data mining algorithm best predicts mobile money transaction growth. This paper was quantitative in nature and used aggregated monthly mobile money data (from Zambian mobile network operators) from 2013 to 2020 as its sample which was collected from Bank of Zambia and Zambia Information Communications and Technology Authority. The paper further used WEKA data mining tool for data analysis following the Cross-Industrial Standard Process for data mining guidelines. The performance from best to least is K-nearest neighbour, random forest, support vector machines, multilayer perceptron and linear regression. The predictions from data mining techniques can be deployed to predict growth of mobile money and hence be used in financial inclusion policy formulation and other strategies that can further improve service delivery by mobile network operators.

Zambia is faced with a lot of challenges; among them is the healthcare delivery. The government has been trying to solve them through establishment of various initiatives which have not helped much; there is need for some smart solutions to be implemented. The growth of big data and wide use of data analytics platforms offer a promise of a more effective and efficient healthcare delivery. This paper aimed at assessing the factors that affect the implementation of big data analytics in the Zambian health sector. Questionnaires and interviews were used to gather information from participants. A quantitative and qualitative approach was used for analysis. The study identified some important factors related to technology, personnel, data management, policies, and electricity. The main contribution that this paper makes is that, it identifies the factors that the Ministry of Health needs to work on in order to effectively implement and use big data analytics.

Spectrum sensing is one of the fundamental functionality performed by a cognitive radio to identify vacant radio spectrum for dynamic spectrum access (DSA). However, there are many challenges still existing before the benefits of DSA can be realized. The challenges include multipath fading, shadowing and hidden primary user (PU) problem. The challenges are more severe in vehicular communication due to unique characteristics such as dynamic topology caused by vehicle mobility. Furthermore, spectrum sensing is dependent on the activities of the PU traffic pattern which are not known in advance. In a typical cognitive radio network, the PU plays a passive role. Therefore, a sensing technique should account for traffic pattern of the PU autonomously. However, most of the proposed spectrum sensing schemes in vehicular communication assumes a static ON/OFF PU model which does not realistically model the PU traffic pattern. In this paper, we propose reinforcement learning (RL) to model the traffic pattern of the PU and use the model to predict channels likely to be free in future. The RL is implemented on road side unit (RSU) which send predicted vacant PU channels to vehicles on the road. Before the channels can be used, vehicles perform spectrum sensing. To account for multipath fading and shadowing, adaptive spectrum sensing is proposed. The results from spectrum sensing, sensing time and PU channel capacity are calculated into a scalar value and used as reward for RL at RSU. The RSU continuously update the reward for channels of interest using sensing history from passing vehicles as reward. Compared to history based schemes from literature, the RL technique proposed in this paper performs better.