Integration of Machine Learning and Optimization Models to Solve Supplier Selection and Order Allocation Problems
Bibliographic record
Abstract
<p dir="ltr">Supplier Selection (SS) and Order Allocation (OA) are two major decision-making processes in supply chain management. Integration of SS and OA is crucial to have effective supply chain designs and efficient operations. In the real world, decision-making processes incorporate complex evaluations of potential suppliers which can be erroneous due to judgemental errors. Moreover, uncertainty in demand can make the decision-making processes difficult. To mitigate these barriers, this dissertation presents hybrid solution approaches that use machine learning techniques and proposes integrating Supplier Selection and Order Allocation (SS&OA) planning frameworks. </p><p dir="ltr">Chapter 2 of this dissertation introduces a two-stage solution approach for SS&OA planning that integrates forecasting techniques and optimization models. Demand forecasting in Stage 1 of this approach addresses the demand ambiguity. Also, Stage 1 considers the demand interdependencies between products that may affect the accuracy of the forecast. Stage 2 develops a Multi-Objective Programming (MOP) model that includes quantitative criteria and the forecasted demands from Stage 1. A Canadian food supply network dataset is considered in this chapter. The results suggest that by integration of inter-product correlations into the forecasting model, businesses can more accurately predict the demand and make better-informed decisions on SS&OA planning when the products are correlated. </p><p dir="ltr">Unprecedented events can disrupt the market dynamics, leading to changes in the balance of supply and demand, and cause fluctuations in historical data. The fluctuations in the data can adversely impact forecasting accuracy. To overcome this challenge during SS&OA planning, Chapter 3 examines a three-stage framework where the effects of data fluctuations are minimized during demand prediction. The framework involves a deep learning technique based on a multistep Long Short Term Memory (LSTM) network for demand prediction. To consider qualitative criteria, a fuzzy Strengths, Weaknesses, Opportunities, and Threats (SWOT) model is developed in Stage 2 of this solution framework. Stage 3 offers the development of a MOP model by retrieving the forecasting model's results from Stage 1 and the fuzzy model's results from Stage 2. The proposed framework's application is also discussed in this chapter using a realistic dataset from the Canadian juice industries. The results show that the fluctuation minimization using machine learning methods may affect the SS&OA planning. It has been also observed that the planning process might be affected by the number of internal and external supplier selection criteria. </p><p dir="ltr">Chapter 4 examines the effect of correlational data components during SS&OA planning. A three stage planning framework is presented in this chapter. Demand forecasting by examining the correlation among data components such as trend, seasonality, and fluctuations involves analyzing the historical data to determine how these components are interrelated. By understanding the relationships between these data components, more precise predictions can be made regarding future demand. In Stage 1 of this planning framework, a modified relational deep learning forecasting method is designed to predict the future demand where the correlations among the data components of different products are considered. To confirm the forecasting accuracy, the proposed deep learning model is compared to a Light-Gradient Boosting Machine and a standalone LSTM. A new fuzzy Principal Component Analysis (PCA) technique is used to calculate the suppliers’ weight in Stage 2. Then, Stage 3 develops a MOP model using Stages 1 and 2 outcomes. Canadian meat sector data is used to discuss the planning framework. The experimental results show that the inter-product correlation functions can change the suppliers and the order quantities during the SSOA planning.</p>
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".