Projects

Identify Diseases in Crops leaf

One of the interesting study fields in agriculture is disease identification from plant photos, for which machine learning principles from computer vision may be utilised. By interfering with photosynthesis, leaf spot diseases harm plants and shrubs. Most leaf spot diseases impact just a tiny fraction of the tree's total leaf area and pose only a minimal threat to the tree's health. Leaf spot infections should be handled seriously if they cause moderate to full leaf loss for the second to fourth year in a row. Leaf loss across multiple growing seasons can lead to decreased development and greater vulnerability to pests and other diseases. Many leaf spot diseases affect a wide variety of native and ornamental plants and shrubs. Many leaf spot diseases have comparable biology and hence therapeutic possibilities. Because of the trickiness of clearly identifying diseases, I decided to present a system for the detection and classification of rice diseases based on the images of infected plants. Plant diseases are one of the factors contributing to the decline in the quality and quantity of crops. Reductions in these characteristics might have a direct impact on a country's total agricultural yield. The main issue is a lack of continuous plant monitoring. Sometimes novice farmers are unaware of illnesses and their incidence periods. In general, diseases can strike any plant at any moment. Continuous monitoring, on the other hand, may help to avoid illness transmission. The diagnosis of plant disease is critical in agriculture. This AI agent seeks to apply Machine Learning and Image Processing techniques to the problem of autonomous disease detection and classification in rice, cassava, cherry, corn, grapes, potato, soybeans, strawberry, and tomato plants. The main outcome of this project is the identification of potential disease outbreaks in farmlands in the early stage of spreading and saving crops from destruction. To achieve that identifier needs a higher accuracy. In this application basically, two models are used to identify the object which has 95.8% accuracy in test data and the most important model which is disease identifier which has 87.6% accuracy also in test data.

Australia weather data analysis and rain prediction

models are binary class, prediction models. widely used algorithms are used to build these models and then use ensembling methods to optimize baseline models to get correct predictions as much as possible. data used to train are feature engineered using imputations, encoding, transformation. all algorithms are coded from scratch using NumPy and then compare with scikit learn models using the same states. feature selection is done using the Pearson correlation coefficient algorithm. full data frame divided into to by 80:20 ratio and 80% of data used to train the model and 20% of data used to testing. evaluations are done through several methods: the number of successful predictions done by each model, ROC curves, and AUC.

Sir Lanka covid-19 situation dashboard

the current situation in the country analysis through various aspects. using interactive visualizations and developed an ARIMA model to forecast new cases and deaths in the next 7 to 28days. get data from OWID covid data repository, local news reports official daily reports of the health ministry then cleaned.

tools used: pandas, Numpy, feature engine, Statmodel, plotly dash, Jupyter notebook, pycharm

House price predictor

using feature engineering techniques increased linear regression model accuracy from 74.6% to 85.91%. choose features using person correlated coefficient algorithm. this model predicts house sale prices in several US cities.

used tools - pandas, SciKit learn, feature engineering, NumPy, SciPy, Jupiter, pycharm, seaborn and plotly for visualizations.

Education

Undergraduate Student

Sabaragamuwa University of Sri Lanka