1 Introduction

This package can also be used to build and/or update NIR applications that are ready to be consumed by the ProxiScout handheld NIR sensors manufactured by BUCHI Labortechnik AG. Once an application is installed in a ProxiScout device, it can be used to predict the properties of a given matrix using the spectral models contained in that application.

This package builds upon the standard structure of the ProxiScout applications, which are conventionally developed with the compiled-executable software “BUCHI Modeller” offered by BUCHI. Therefore, the files output by proximetricsR follow the same structure of the ones output by “BUCHI Modeller”. No changes or improvements on these output files have been conducted for the development of proximetricsR.

ProxiScout applications differ from ProxiMate applications in their internal file format: a ProxiScout application consists of two JSON files (operations.json and model_info.json) that are uploaded to the NeoSpectra Portal and subsequently synchronised to the ScanApp Mobile Application, where predictions are executed locally on the mobile device.

2 ProxiScout predictive application package

2.1 Overview

A ProxiScout predictive application is deployed through the NeoSpectra ecosystem and consists of two JSON files:

• Operations.json: Defines the ordered modeling workflow, including preprocessing, variable selection, and predictive model execution.

• Model_Info.json: Contains model metadata, performance statistics, and application configuration settings.

Together, these files define the predictive workflow and configuration required for deployment. The application package is uploaded to the NeoSpectra Portal and synchronized to the ScanApp Mobile Application, where predictions are executed locally on the mobile device.

2.2 Deployment workflow

1. The predictive application package is generated as:

   • operations.json

   • model_info.json

2. The files are uploaded to the NeoSpectra Portal.

3. The application is synchronized to the ScanApp Mobile Application.

4. ScanApp downloads the application configuration and executes the predictive workflow during sample analysis.

3 operations.json

3.1 Purpose

The operations.json file defines the complete modeling pipeline used to transform raw spectral measurements into prediction results.

The pipeline consists of a sequence of operations executed in a predefined order. These operations may represent spectral preprocessing, variable selection, feature extraction, scaling, regression, classification, or any other modeling step required by the predictive application.

3.2 File structure

The file contains an ordered collection of operation objects.

3.2.1 Operation Object

Each operation is represented by a JSON object containing the following fields:

3.2.2 Example

{
    "id": 83,
    "params": [
        9.0,
        3.0,
        1.0
    ],
    "index": 5
}

3.3 Execution model

Operations are executed sequentially according to the value of the index field.

Each operation receives the output of the previous operation as input and produces output that is passed to the next operation in the pipeline.

The modeling workflow may include operations for:

• Spectral preprocessing
• Variable selection
• Data transformation and scaling
• Regression or classification
• Post-processing of prediction results

3.4 proximetricsR function reference

When using proximetricsR to build ProxiScout applications, preprocessing steps are automatically serialized to the correct JSON format. This reference maps proximetricsR preprocessing functions to their corresponding ProxiScout operation IDs and JSON structure:

3.4.1 Spectra scale (ID: 37)

• proximetricsR: None (spectra must be provided in reflectance, 0–100%)

• Requires params: Yes (scale factor)

{
    "id": 37,
    "params": 0.01,
    "index": 0
}

3.4.2 Get absorbance (ID: 29)

• proximetricsR: prep_transform()

• Requires params: No

{
    "id": 29,
    "params": [],
    "index": 5
}

3.4.3 Average readings (ID: 7)

• proximetricsR: None

• Requires params: No

{
    "id": 7,
    "params": [],
    "index": 12
}

3.4.4 SNV (ID: 2)

• proximetricsR: prep_snv()

• Requires params: No

{
    "id": 2,
    "params": [],
    "index": 0
}

3.4.5 Detrending (ID: 3)

• proximetricsR: prep_detrend()

• Requires params: Yes (detrending degree)

{
    "id": 3,
    "params": [2],
    "index": 3
}

3.4.6 Savitzky-Golay smoothing and differentiation (ID: 83)

• proximetricsR: prep_derivative(algorithm = "savitzky-golay") or prep_smooth(algorithm = "savitzky-golay")

• Requires params: Yes (window length, polynomial order, derivative order, coefficients)

{
    "id": 83,
    "params": [
        9.0,
        7.0,
        1.0,
        105.0,
        1.0,
        0.0,
        -0.0035,
        0.038,
        -0.20,
        0.8,
        8.22e-15,
        -0.8,
        0.2,
        -0.038,
        0.0035
    ],
    "index": 5
}

3.4.7 Variable selection (ID: 17)

• proximetricsR: prep_wav_trim()

• Requires params: Yes (selected variable indices)

{
    "id": 17,
    "params": [
        17,
        18,
        19,
        200,
        201,
        202,
        203
    ],
    "index": 12
}

4 model_info.json

4.1 Purpose

The model_info.json file contains metadata associated with the predictive model and application configuration.

This information is used by ScanApp to display model information, configure measurement acquisition settings, and provide model performance statistics.

4.2 Contents

The file contains:

• Model performance metrics for calibration, cross-validation, and testing
• Prediction range information
• Dataset statistics used during model development
• Measurement and prediction averaging settings
• Application execution configuration

4.2.1 Example

{
    "executionOrder": 0,
    "RMSECalib": 0.58,
    "R2Calib": 0.814,
    "RMSECV": 0.701,
    "R2CV": 0.73,
    "BiasCV": 0.013,
    "RPDCV": 1.925,
    "RMSETest": 0.598,
    "R2Test": 0.947,
    "BiasTest": 0.145,
    "RPDTest": 2.804,
    "avgReadings": 1,
    "avgPredictions": 1,
    "minValue": 3.1,
    "maxValue": 9.5,
    "NumberOfSamples": 119,
    "NumberOfMeasurements": 6
}