LightGBM Property Price Predictor
Budget: $30 – $250 USD
I have an Excel file containing roughly 14,000 rows of historical property sales. The columns already include property size, location, number of rooms and several additional attributes. What I need is a well-tuned LightGBM regression model that can learn from this data and, once trained, can be queried through a very simple web form.
Here is the flow I’m expecting:
• Data handling: you’ll clean any obvious outliers or missing values, encode categorical fields (location, for example) and split the data into train/validation sets so I can see an honest performance report.
• Model building: use LightGBM, run reasonable hyper-parameter tuning and share the metrics you achieve (RMSE or MAE are fine). Keep the code clearly commented so I can tweak it later.
• Deployment layer: a lightweight web form (Flask or Streamlit are both fine) that asks for size, location, number of rooms and the extra features present in the file, then returns the predicted price instantly. No heavy styling is needed—function over aesthetics.
• Handover: please provide the cleaned dataset (or a script that performs the cleaning on load), the trained model file, the full source code for the web form, and brief instructions for running everything locally (python version, libraries, and the command to start the server).
If anything in the data requires special handling, let me know early so we can adjust. Looking forward to a straightforward, accurate predictor that I can demonstrate to stakeholders with a single click.
Here is the flow I’m expecting:
• Data handling: you’ll clean any obvious outliers or missing values, encode categorical fields (location, for example) and split the data into train/validation sets so I can see an honest performance report.
• Model building: use LightGBM, run reasonable hyper-parameter tuning and share the metrics you achieve (RMSE or MAE are fine). Keep the code clearly commented so I can tweak it later.
• Deployment layer: a lightweight web form (Flask or Streamlit are both fine) that asks for size, location, number of rooms and the extra features present in the file, then returns the predicted price instantly. No heavy styling is needed—function over aesthetics.
• Handover: please provide the cleaned dataset (or a script that performs the cleaning on load), the trained model file, the full source code for the web form, and brief instructions for running everything locally (python version, libraries, and the command to start the server).
If anything in the data requires special handling, let me know early so we can adjust. Looking forward to a straightforward, accurate predictor that I can demonstrate to stakeholders with a single click.