Abstract:
Environmental factors play a major role in tea growing and plucking stages
and these factors must have to be within the favourable range to get quality tea
production. At present tea, pluckers cannot identify the exact duration for tea plucking
and they do not have sources to identify and pick tea leaves from tea buckets without
overflowing which can cause physical damages to tea leaves. This research addresses
the above issues by creating data forecasting models that provide significant guidance
to make decisions in many areas especially in tea cultivation, plucking, and
transportation. Three devices were developed to capture real-time weather data namely
soil PH, surface temperature, and Humidity. Above sensors data were transmitted over
GPRS using a GSM module. Evaluated results of datasets with actual data values and
analysed with different prediction algorithms such as Voted perceptron, Decision table
algorithm, Multilayer perceptron, and Simple linear regression. After observing all the
aspects, several variables, and prediction accuracy for data samples, the most relevant
algorithm to build the prediction models were decided. The models were executed with
a different combination of factors and analysed the output prediction result to sort the
most accurate factor combination for the dataset. Models were built to predict the most
suitable periods having optimum environmental conditions to pluck tea leaves,
production forecasts by considering environmental and soil conditions, and transport
scheduling for plucked tea leaves before quantity overflows. Above mention, the
model was helped to schedule the plucking process while enhancing the quality of tea
leaves. Further, this study introduced the smart tea plucking basket to control the realtime weather conditions and reduce human malpractices while maintaining optimum
quality. The research recommends assessing this model with different algorithms to
fine-tune the performance and to build a general model that can be applied when
enhancing other quality factors.
Keywords: Internet of Things (IOT), Voted perceptron, Decision table algorithm,
Multilayer perceptron, Simple linear regression
