Python machine learning predictive models for real estate sales
Budget: $250 – $750 USD
I need a python and MapGIS pro to help me with my CSV Xcel file dataset. I need 3 predictive models that have extremely high accuracy for 4 target variables and provide all of the code with markdowns explaining. Make sure you eliminate variables with low influence and any collinearity, standardize variables of different formats and data types and prevent over-fitting. Optimize the models by adjusting epochs and other items. Account for null values, construct flag variables for categoricals, Test for accuracy and select the most accurate models. Try several model types to find the best one. Construct the 3 models with the preprocessed data. Each of the 3 models is a different model type. The new models will need to learn daily as new data is updated. The output should be a column on the unknown dataset with the populated target variables.
Model 1: Needs to be a model that predicts the future closing price.
Model 2: Alternative model that predicts the future closing price.
Model 3: I need a matching mechanism. I have found that the Euclidean distance may be a great way to do this. You will need to standardize the variables into a scaler. When finished I should be able to enter a new record and the model can match the new data to the most similar records of the sales data. Then we would need to average the top 3, 5, 10, 25 sales prices of the closest sales.
Data Gathering: The size of the dataset is between 50,000 and 100,000 records so it is not very large. You will need to run a text mining script to obtain some variables from the description of the property (description is also in the excel file).
Finally, I need help with mapping these properties out with their addresses and placing them into categories based on their geographic location: i.e. Subdivisions.
Deliverables:
Python script using Jupiter Notebook. 3 models for 4 target variables. This product section is satisfied when I am able to run the model without any issues and the model accurately produces predicted target variables. The script will have all of the data wrangling and preprocessing and accuracy testing and preventative measures for over-fitting.
Provide full documentation of the models because I am sure later on I will add more data in the future to improve model performance. I will need to know why you used the codes in each step as well which you can place in markdown cells.
I need an embed code to put it into my website:
Essentially on the webpage you would enter the unique identifier and the output would be the 4 target variables.
The same would happen for the matching mechanisms.
So once the project is done, I will need to be able to access it and use it on my website. The dataset will be stored on DropBox.
Provide up to 90 days or more of prompt technical support for any questions I may have after the project is finished.
Model 1: Needs to be a model that predicts the future closing price.
Model 2: Alternative model that predicts the future closing price.
Model 3: I need a matching mechanism. I have found that the Euclidean distance may be a great way to do this. You will need to standardize the variables into a scaler. When finished I should be able to enter a new record and the model can match the new data to the most similar records of the sales data. Then we would need to average the top 3, 5, 10, 25 sales prices of the closest sales.
Data Gathering: The size of the dataset is between 50,000 and 100,000 records so it is not very large. You will need to run a text mining script to obtain some variables from the description of the property (description is also in the excel file).
Finally, I need help with mapping these properties out with their addresses and placing them into categories based on their geographic location: i.e. Subdivisions.
Deliverables:
Python script using Jupiter Notebook. 3 models for 4 target variables. This product section is satisfied when I am able to run the model without any issues and the model accurately produces predicted target variables. The script will have all of the data wrangling and preprocessing and accuracy testing and preventative measures for over-fitting.
Provide full documentation of the models because I am sure later on I will add more data in the future to improve model performance. I will need to know why you used the codes in each step as well which you can place in markdown cells.
I need an embed code to put it into my website:
Essentially on the webpage you would enter the unique identifier and the output would be the 4 target variables.
The same would happen for the matching mechanisms.
So once the project is done, I will need to be able to access it and use it on my website. The dataset will be stored on DropBox.
Provide up to 90 days or more of prompt technical support for any questions I may have after the project is finished.