Machine Learning Pricing Recommendations Without Loss Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional pricing recommendation systems face issues with data sparsity and inaccurate predictions due to data segmentation, leading to misleading results and missed macro-level pricing effects, especially when loss data is unavailable.

Innovation Solution

A machine learning approach that trains a logistic model on unsegmented historical transaction data using an artificial neural network, generating pricing recommendations and win rate curves without relying on loss data, allowing broader insights across various types of transactions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If historical transaction data is segmented into smaller subsets for analysis, then the analysis can focus on specific customer and product segments, but data sparsity occurs leading to inaccurate predictions and misleading results

Engineering Contradiction:

Improvepricing recommendation accuracyVSAvoiddata availability

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent merges segmented transaction data back into a unified dataset for training the logistic model. Instead of analyzing segmented data separately which causes data sparsity, the system combines all historical transaction data to maintain sufficient data volume while still capturing segment-specific patterns through the model's ability to learn from heterogeneous data.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The logistic model is designed to be universal across different customer segments and product types. Rather than creating separate models for each segment, a single logistic model processes all transaction data, enabling it to learn general pricing patterns while adapting to specific segment characteristics, thus avoiding data sparsity in individual segments.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Measurement precision

If conventional pricing systems use segmented data, then specific segment analysis is improved, but macro-level pricing effects are missed and predictions become inaccurate

Engineering Contradiction:

Improvepricing prediction accuracyVSAvoidmacro-level pricing effects

Core Design Contradiction:

Measurement precisionVSLoss of information

Solution Approach 1:

The system merges segmented transaction data into a comprehensive dataset that preserves macro-level pricing effects. By training the logistic model on unified historical transaction data rather than isolated segments, the model captures both segment-specific and macro-level pricing patterns, preventing information loss about overall market trends.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The logistic model transforms the analysis by operating in a higher-dimensional feature space that incorporates both segment-specific attributes and macro-level pricing factors. This dimensional transformation allows the model to simultaneously consider detailed segment characteristics and broader market trends without being constrained by data segmentation.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

3Reliability

If loss data is used for training pricing models, then win rate calculations can be performed, but the system becomes dependent on available loss data which may not always be present

Engineering Contradiction:

Improvewin rate calculationVSAvoidapplicability without loss data

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The logistic model performs self-service by generating its own training signals from historical transaction data without requiring external loss data. The model learns to predict pricing outcomes by identifying patterns in successful transactions and inferred pricing dynamics, making the system self-sufficient and adaptable to environments where loss data is unavailable.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

Instead of requiring loss data to train the model (conventional approach), the patent inverts the approach by training the logistic model on available historical transaction data and using it to infer win rates and pricing recommendations. This inversion makes the system adaptable to situations where loss data is missing by deriving necessary information from alternative data sources.

Inventive Principle:

Principle #13The other way round (Inversion)

Data Source

PatentUS20250238830A1Machine learning techniques for generating recommendations for a transaction without loss data

Publication Date: 2025.07.24 CONGA CORPORATION

AI summary

Techniques for generating a predicted win rate curve without using loss data are disclosed. An example method includes receiving historical transaction data comprising a plurality of transactions, each transaction comprising a plurality of attributes. The method also includes processing the historical transaction data to generate training data comprising features extracted from the plurality of attributes. The method also includes training, by a processing device, a logistic model using the training data and a predicted price. Training the logistic model includes providing the predicted price and a subset of the features at an input layer of a neural network and training the neural network to generate a mapping from the predicted price and the subset of the features to parameters of a predicted win rate curve generated at an output layer of the neural network.