Abstract
Internet of things (IoT) produces massive data from devices embedded with sensors. The IoT data allows creating profitable services using machine learning. However, previous research does not address the problem of optimal pricing and bundling of machine learning-based IoT services. In this paper, we define the data value and service quality from a machine learning perspective. We present an IoT market model which consists of data vendors selling data to service providers, and service providers offering IoT services to customers. Then, we introduce optimal pricing schemes for the standalone and bundled selling of IoT services. In standalone service sales, the service provider optimizes the size of bought data and service subscription fee to maximize its profit. For service bundles, the subscription fee and data sizes of the grouped IoT services are optimized to maximize the total profit of cooperative service providers. We show that bundling IoT services maximizes the profit of service providers compared to the standalone selling. For profit sharing of bundled services, we apply the concepts of core and Shapley solutions from cooperative game theory as efficient and fair allocations of payoffs among the cooperative service providers in the bundling coalition.
Introduction
The modern Internet of Things (IoT) landscape is defined by a "Data-to-ML-to-Service" pipeline. While technical advancements in sensors and model accuracy are frequently discussed, the economic framework governing these services is often overlooked. This paper provides a rigorous mathematical model to determine optimal pricing for machine learning-based IoT services, treating service quality as a direct function of training data volume.
The IoT Market Ecosystem
The authors define a three-tier market structure that bridges the gap between raw data generation and consumer value. In this ecosystem, the service provider acts as a central intermediary:
- Data Vendors: These entities collect and sell raw sensor data from various sources like smart homes, healthcare IoT, and industrial sensors.
- Service Providers: They purchase this data to train machine learning models, transforming raw information into "intelligence" (e.g., predictive maintenance or context-aware alerts).
- Customers: End-users who pay a subscription fee based on the perceived value and quality of the resulting service.
The core economic challenge lies in the fact that service quality is dynamic. In machine learning, accuracy depends on the volume of training data. A provider must balance the cost of purchasing data against the potential revenue from subscription fees, which are themselves sensitive to service quality.
Modeling Service Quality and Pricing
To solve this optimization puzzle, the research introduces a quality function ($q$) modeled as an exponential function: $q(n) = \alpha_1 - \alpha_2 \exp(-\alpha_3 n)$, where $n$ represents the data size. This captures the law of diminishing returns inherent in machine learning: initial data samples provide significant accuracy gains, while subsequent samples offer marginal improvements.
By formulating non-linear optimization problems, the authors derive closed-form solutions. These allow providers to calculate optimal prices and data requirements instantaneously, avoiding the need for computationally expensive iterative simulations.
Bundling and Cooperative Strategy
A key insight of the paper is the economic superiority of service bundling. When multiple providers cooperate—for instance, bundling smart home security with energy management—the total profit of the coalition typically exceeds the sum of individual standalone profits.
To ensure stability and fairness within these coalitions, the authors employ cooperative game theory:
- The Core Solution: Ensures the partnership is stable and no member has an incentive to leave.
- The Shapley Value: Provides a mathematically "fair" profit split based on each provider's marginal contribution to the bundle's overall value, preventing "free-riding."
Empirical Validation
The models were validated using four diverse real-world datasets: 20 Newsgroups (text categorization), Sentiment140 (sentiment analysis), MNIST (OCR), and a custom Google Survey of 100 users to model willingness-to-pay.
Numerical results confirmed that the proposed exponential quality function fits various ML algorithms, including Deep Learning and SVMs. Furthermore, the study identified specific "data saturation points" beyond which purchasing additional data becomes economically irrational, despite minor accuracy gains. For example, numerical results showed that a bundled service generated a total profit significantly higher ($F_K \approx 19.67$) than the sum of standalone profits ($17.89$).
Conclusion
This research provides a foundational economic framework for the IoT industry. By linking machine learning performance metrics directly to market pricing strategies, it offers actionable insights for service providers. Future work may expand these models to account for competitive price wars and non-uniform distributions of consumer willingness-to-pay.
Summary updated