Inspiration My father, father-in-law and grandfather have been farmers, and their dedication to agriculture has always inspired me. This personal connection motivated me to choose a project that could directly support them and others in the farming community. I aim to develop tools and solutions that can assist farmers in making more informed decisions, improving crop yields, and enhancing their farming practices.
What it does We have developed a webpage that integrates a machine learning model. When users input details such as humidity, temperature, salinity, soil type, and soil texture, the model processes this information and provides crop recommendations. For example, the user might receive suggestions like "Apple" for fruits or "Lettuce" for vegetables based on the provided conditions.
How we built it *We started by extracting data from the .db file and converting it into a DataFrame. After preprocessing the data, we removed unnecessary columns such as ID fields and descriptions to streamline the model-building process. We applied techniques like ordinal encoding, target encoding, and scaling to prepare the data.
Next, we built a neural network model and tuned it using grid search. We also explored ensembling techniques by combining neural networks with Random Forest and K-Nearest Neighbors classifiers. Ultimately, the neural network model provided the best results.
Once the model was trained and optimized, we saved it for deployment. To build the web application, we used Flask for the backend and HTML, CSS, and JavaScript for the frontend.*
Challenges we ran into *We faced several challenges during the project. One of the main hurdles was connecting Python to the front-end dashboard seamlessly. Integrating the machine learning model with the web application required careful handling of data and ensuring smooth communication between the backend and frontend.
Another challenge was identifying the top contributors to the machine learning algorithm's output. We needed to determine which features had the most significant impact on the predictions, which involved extensive analysis and feature selection.
Lastly, during the preprocessing phase, we had to carefully figure out the most relevant features to include in the model. Balancing data cleaning, encoding, and scaling while preserving essential information was a critical step to building an effective model.*
Accomplishments that we're proud of One of our proudest achievements is successfully connecting the front end with the backend, ensuring that the machine learning model operates seamlessly behind the scenes. This integration allows the user to interact with the web application without needing to understand the complexities of the backend. The system provides clear and actionable results, making it accessible for a layman to easily obtain the crop recommendations they need.
What we learned We gained valuable experience connecting and analyzing data from the .db file, performing preprocessing, and removing irrelevant data to improve the quality of our dataset. We experimented with multiple machine learning algorithms to optimize accuracy, learning how different approaches impact model performance. Additionally, we learned the intricacies of integrating the backend with the front end, ensuring smooth data flow and delivering user-friendly results.
What's next for FalconX *Moving forward, we can plan to enhance the model’s capabilities by incorporating more diverse datasets, such as weather patterns and crop-specific growth data, to further improve prediction accuracy. We can also aim to expand the web application with additional features, such as personalized crop recommendations based on geographic location and seasonal trends. Additionally, integrating real-time data inputs from IoT devices on farms could provide more dynamic recommendations and support farmers in making data-driven decisions throughout the growing season.
We can also plan to continuously refine the machine learning models, explore advanced algorithms, and further optimize the user experience to ensure the application remains a valuable tool for farmers.
Log in or sign up for Devpost to join the conversation.