Crop Yield Prediction Using Bidirectional Lstm With Real-Time Data Visualization

Abstract

Crop yield prediction plays a pivotal role in ensuring
food security and optimizing agricultural planning.
Traditional methods, such as statistical regression
or simple machine learning algorithms, often fail to
account for the nonlinear and temporal nature of
farming data. In this project, we introduce an
advanced deep learning approach using
Bidirectional Long Short-Term Memory (BiLSTM)
networks to enhance the accuracy of crop yield
prediction. The BiLSTM model is trained on a
combination of historical weather data, soil metrics,
and agricultural practices, enabling it to learn both
past and future contextual relationships in timeseries
data. To make the model practical and usercentric,
we integrate a real-time data visualization
dashboard that continuously updates yield
predictions as new environmental data streams in.
The system aims to assist farmers, agronomists, and
policymakers in making proactive decisions that can
significantly improve productivity and
sustainability.
Our system, titled "Milifare," integrates a real-time
data visualization interface to assist farmers,
researchers, and policymakers in decision-making.
The model is trained on multivariate time-series
data including weather patterns, soil parameters,
and historical crop yields. Experimental results
show improved prediction accuracy over traditional
methods. This research offers a comprehensive
platform combining predictive analytics with an
intuitive visualization dashboard.