Abstract:The effectiveness of neural network based models in forecasting daily precipitation, based on ground level measurements obtained from a cluster of weather stations in the dry zone of Sri Lanka, is presented. The implemented networks were based on a feed-forward back-propagation technique. A cluster of ten neighbouring weather stations having 30 years of daily precipitation data (1970 -1999) was used in training and testing the models. Twenty years of daily precipitation data were used to train the networks while ten years of daily precipitation data were used to test the effectiveness of the models. One model was developed to forecast the precipitation occurrences such as 'rain' or 'no rain', while another model was developed to predict the amount of precipitation at several sub levels using fuzzy techniques. Overall, the models were able to predict the occurrence of daily precipitation with an accuracy of 79±3%. Only the nearest neighbours contributed to improving the accuracy of predictions. In the dry zone, the accuracy of predicting the dry days was superior compared to predicting wet days except during the rainy season. Fuzzy classification produced a higher accuracy in predicting 'trace' precipitation than other categories.

Abstract-In Sri Lanka (SL), graduands' employability remains a national issue due to the increasing number of graduates produced by higher education institutions each year. Thus, predicting the employability of university graduands can mitigate this issue since graduands can identify what qualifications or skills they need to strengthen up in order to find a job of their desired field with a good salary, before they complete the degree. The main objective of the study is to discover the plausibility of applying machine learning approach efficiently and effectively towards predicting the employability and related context of university graduands in Sri Lanka by proposing an architectural framework which consists of four modules; employment status prediction, job salary prediction, job field prediction and job relevance prediction of graduands while also comparing performance of classification algorithms under each prediction module. Series of machine learning algorithms such as C4.5, Naïve Bayes and AODE have been experimented on the Graduand Employment Census -2014 data. A pre-processing step is proposed to overcome challenges embedded in graduand employability data and a feature selection process is proposed in order to reduce computational complexity. Additionally, parameter tuning is also done to get the most optimized parameters. More importantly, this study utilizes several types of Sampling (Oversampling, Undersampling) and Ensemble (Bagging, Boosting, RF) techniques as well as a newly proposed hybrid approach to overcome the limitations caused by the class imbalance phenomena. For the validation purposes, a wide range of evaluation measures was used to analyze the effectiveness of applying classification algorithms and class imbalance mitigation techniques on the dataset. The experimented results indicated that RandomForest has recorded the highest classification performance for 3 modules, achieving the selected best predictive models under hybrid approach having an area under the ROC curve interpretation as an 'Excellent' experiment, while a C4.5 Decision Tree model under Ensemble approach has been selected as the best model of the remaining module (Salary Prediction module).

No abstract

No abstract

The performance of artificial neural networks in forecasting short range (3-6 hourly) occurrence of rainfall is presented. Feature sets extracted from both surface level weather parameters and satellite images were used in developing the networks. The study was limited to forecasting the weather over Colombo (79°52' E, 6°54' N), the capital of Sri Lanka. From the available ground level weather parameters, a total of seven parameters, namely, pressure, temperature, dew point, wind direction, wind speed, cloud amount and rainfall were selected for the present study. From satellite images, four types of images viz., visible image of clouds, infrared image of clouds, infrared colour image of clouds and water vapour image of clouds were used. The best performance was observed for hybrid models that combine ground level and satellite observations, with 75% accuracy for short range forecasting. A strong seasonal dependence in the accuracy of forecasting linked to monsoons was observed.