OPTMOVE Clustering Algorithm for SKU Forecasting Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Retailers face challenges in accurately forecasting sales for stock keeping units (SKUs) due to incomplete or erratic sales data, especially in e-commerce settings, where multivariate models are needed to forecast groups of items together, but existing clustering methods are inefficient and prone to local minima, leading to suboptimal grouping and forecasting.
Innovation Solution
The OPTMOVE clustering algorithm, which sets a maximum number of vertices per cluster and moves vertices to improve cluster quality, using a randomized approach to prevent outlier influence and ensure efficient grouping, allowing for better forecasting by minimizing dissimilarity within clusters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing clustering methods like K-means are used to group SKUs for forecasting, then the forecasting process can be performed on groups of items, but the clustering results are prone to local minima and produce suboptimal grouping
Solution Approach 1:
The patent applies dynamics by making the clustering process adaptive and iterative. The system dynamically adjusts cluster assignments through multiple passes, allowing clusters to evolve and improve over time rather than being static. This is achieved by repeatedly assigning SKUs to clusters based on similarity metrics and refining the groupings across iterations to escape local minima and achieve better forecasting accuracy.
Solution Approach 2:
The patent implements feedback by using forecasting performance as a criterion to evaluate and refine cluster quality. The system measures the impact of clustering on forecasting accuracy and uses this feedback to adjust and improve cluster assignments in subsequent iterations, creating a closed-loop system that continuously optimizes both clustering and forecasting.
2Reliability
If multivariate models are used to forecast groups of items together, then forecasting can be performed for items with insufficient individual sales data, but the quality of forecasts depends significantly on how groups are formed
Solution Approach 1:
The patent applies parameter changes by utilizing multiple dimensions or parameters for grouping SKUs beyond traditional categories. The system considers various parameters such as sales patterns, seasonality, product attributes, and other relevant features to create more nuanced and effective clusters, thereby improving forecasting capability while managing complexity through systematic parameter evaluation.
3Ease of manufacture
If traditional clustering algorithms are used, then the implementation is straightforward, but they are inefficient and prone to getting stuck in local minima
Solution Approach 1:
The patent applies preliminary action by performing initial clustering and then systematically refining the results through multiple passes and optimization steps. Rather than relying on a single straightforward algorithm, the system prepares initial groupings and then iteratively improves them, escaping local minima through repeated refinement while maintaining implementation feasibility.
Data Source
AI summary
A system and method for grouping units for forecasting purposes is presented. A plurality of stock keeping units (SKUs) is presented to an embodiment. Initial medoids are chosen based on a vertex within a set of vertices, each of which represent a SKU. Then, each vertex within the set of vertices is associated with its closest medoid to form initial clusters. There can be a cap on the number of vertices in each cluster. Thereafter, an iterative algorithm is performed wherein a probability is assigned to each vertex. One or more vertices are randomly chosen, with the weights of the vertices weighting the random choice. The chosen one or more vertices are moved to another cluster. The algorithm is performed until no further improvements result from moving one or more vertices to another cluster. Other embodiments are also disclosed herein.