In the context of digitalization and stock competition in consumer finance, the loss of high-value credit card customers (20% contributing 80% of profits) directly affects corporate profits. The traditional RFM model has a single dimension and collinearity, making it difficult to capture behavioral changes; K-means clustering has difficulties in determining K values and dynamic adaptability; Static machine learning models are not easily able to reflect complex factors such as economic fluctuations. This study integrates transaction frequency, amount, interaction behavior, transaction changes, and credit characteristics to construct an improved RFM+CBC model that breaks through traditional dimensional limitations. Optimize K-means clustering by combining contour coefficient and Calinski Harabasz index, determine indicator weights using entropy weight method, and achieve precise customer segmentation. Based on logistic regression, random forest, XGBoost, and Stacking fusion models for churn prediction, the results show that customer value segmentation significantly improves prediction performance. XGBoost has an accuracy rate of 97.73%, a recall rate of 93.10%, and a Stacking accuracy rate of 97.98%among high-value customer groups, verifying the core value of segmentation in predicting high-value customers. In the future, it is possible to expand multi-source data (such as social media behavior), develop dynamic segmentation and real-time prediction systems, and apply deep learning technology to further strengthen the integration and application of artificial intelligence in marketing data analysis and business decision-making, enhancing customer loyalty and long-term benefits.

Yiyu Yang. The Integration and Application of Ai Technology in Marketing Data Analysis and Business Decision-Making. International Journal of Business Management and Economics and Trade (2026), Vol. 7, Issue 2: 11-18. https://doi.org/10.38007/IJBMET.2026.070202.

[1] Giannini M. Focus your resources with Pareto principle[J].Automotive news, 2025 (TN.7188): 100.

[2] Zheng Y, Gan W, Chen Z, et al. SeqRFM: Fast RFM analysis in sequence data[J]. Information Sciences, 2026, 723(c):122652.

[3] Lai H, Huang T, Lu B L, et al. Silhouette coefficient-based weighting k-means algorithm[J]. Neural Computing and Applications, 2025, 37(5):3061-3075.

[4] Paulraj K, Murraliraj R, Kumar S S M, et al. Sales price prediction with customer segmentation[J]. AIP Conference Proceedings, 2025, 3279(1):020008.

[5] Rubinstein M A. Reimagining the airport-customer relationship through customer relationship management[J] .Journal of Airport Management, 2025, 20(1)..

[6] Park H, Bai S, Ye Z, et al. Graph Canonical Correlation Analysis[J]. 2025.

[7] Gan G. A k-Means Algorithm with Automatic Outlier Detection[J]. ELECTRONICS, 2025, 14(9):1723.

[8] Kumari B, Vinaya Prasad T. Customer Churn Prediction using Data Science[J]. International Scientific Journal of Engineering & Management, 2025, 4(7).

[9] Bosso V D A, Nardini R M, De Sousa M A D A, et al. An Area-Efficient and Low-Error FPGA-Based Sigmoid Function Approximation[J]. Applied Sciences (2076-3417), 2025, 15(21).

[10] Ulitin N V, Shadrina G R, Anisimova V I, et al. Interpretation of the Structure-Glass Transition Temperature Relationship for Organic Homopolymers with the Use of Increment, Random Forest, and Density Functional Theory Methods[J].Journal of Structural Chemistry, 2025, 66(5).