Auto generate table input type for predictFcn exported model

Hi

I have already read the documentaion about the methods requiervedVariable and hotToPredict of the exported model.

I also manually created the required input format for predictFcn.

My dataset is a table with lots of categorical and double data. I select a few of the columns as feature in regression lerarner app.

I am looking for an easier way to create a input data table for testing and predicting with my model after exporing. Its a little hard to each time create a table manually with the variable names and variable types.

This is my Method now:

if trained == true

load trainedModel.mat

varNames = ["BlockTime","FlightTime","dayOfYear","hourOfDay","payloadKg","totalWeightKg"];

varTypes = ["double","double","double","double","double","double"];

inT = table('Size',size(varNames),'VariableTypes',varTypes,'VariableNames',varNames);

intT(1,:) = array2table([BlockTime FlightTime dayofYear PayloadKG TotalKG]);

fuelPredict = trainedModel.predictFcn(inT)

end

but my main data has lots of other columns so I cant use that for input template

opts.Sheet = "Sheet1";

opts.DataRange = "A2:AM5115";

% Specify column names and types

opts.VariableNames = ["ACType", "ACReg", "FlightDateM", "FlightNo", "Origin", "Destination", "Airborn", "OffBlock", "TouchDown", "OnBlock", "BaggageKG", "BaggagePD", "AdultCount", "InfantCount", "ChildCount", "FlightTime", "BlockTime", "SitaFlightTime", "RemainingFuelKG", "RemainingFuelLt", "RemainingFuelPD", "UpliftFuelKG", "UpliftFuelLT", "UpliftFuelPD", "ArrFuelKG", "ArrFuelLT", "ArrFuelPD", "DiffFuelKG", "DiffFuelLT", "DiffFuelPD", "RampFuelKG", "RampFuelLT", "RampFuelPD", "SitaFuelKG", "SitaFuelLT", "SitaFuelPD", "TaxiFuelKG", "TaxiFuelLT", "TaxiFuelPD"];

opts.VariableTypes = ["categorical", "categorical", "datetime", "string", "categorical", "categorical", "datetime", "datetime", "datetime", "datetime", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double", "double"];

[Siddharth Bhutiya]

Can you share a sample MAT file of what the initial data looks like and what you want the final table to look like ?

Answer 1

[Alireza Ghaderi]

I wrote this function for myself once for ever:)

just share it here if anyone need it:

function prediction = smartPredict(model, inputTable)
% smartPredict Predict the output using the given model and input table
%
% Syntax: prediction = smartPredict(model, inputTable)
%
% Inputs:
%   model: A regression learner model created by the Regression Learner App
%   inputTable: A table containing the input data, which may or may not have
%               all the required variables
%
% Outputs:
%   prediction: A scalar or vector containing the predicted output(s)
%               based on the given input data
%
% Example:
%   mdl = fitlm(X, y); % Create a linear regression model
%   input_data = table(X1, X2); % Create an input table
%   pred = smartPredict(mdl, input_data); % Get the prediction
%
% Notes:
%   - This function will report missing required variables and ignore
%     any extra columns in the input table.
%   - The output is a scalar if the input data has one row and a vector if
%     the input data has multiple rows.

    % Get the required variables from the model
    requiredVariables = model.RequiredVariables;

    % Check if the input table contains all the required variables
    missingVariables = setdiff(requiredVariables, inputTable.Properties.VariableNames);

    if ~isempty(missingVariables)
        % Report missing variables
        error('smartPredict:missingVariables', ...
              'The input table is missing the following required variables: %s', ...
              strjoin(missingVariables, ', '));
    end

    % Select only the required variables from the input table
    selectedInputTable = inputTable(:, requiredVariables);

    % Make the prediction
    try
        prediction = model.predictFcn(selectedInputTable);
    catch e
        % Report errors during prediction
        error('smartPredict:predictionError', ...
              'An error occurred during prediction: %s', e.message);
    end
end