Last two decades there has been a tremendous shift in technology. We are seeing AI/ML become the breakthrough technology and most anticipated to have a transformative impact on most of the industries.
Data as such has become the most important asset which when leveraged meaningfully can bring in significant value-add to businesses.
In our use case we picked up Iris dataset – this can help avoid risks and help Pharmaceutical companies as well as Floriculture Industries.
A short brief:
There are three species in iris flower dataset - Sentosa, Versicolor, and Virginica. Although these species appear similar, they have different traits. And, some varieties of the iris flower are poisonous, the medical industry occasionally needs to identify the species in order to manufacture skin products.
Therefore, we require a model to accurately identify poisonous flower species.
H2o.ai was chosen as a platform along with R programming language.
To implement this flower species detection various models were checked from a feasibility perspective. Further scalability and ease of integration factors were also considered.
With brainstorming session logical flow was arrived at and few conclusions were derived. Data was split into training and test data sets. Libraries of h2o were imported in R. Post that the cluster was initiated. The model was trained and validated.
Based on analysis, random forest model was adjudged as the most suitable model with higher accuracy of prediction. With new data values into the trained model, predictions were pretty good to isolate the species categorically as poisonous or medicinal.
