| MACHINE LEARNING as a tool to improve the effectiveness of marketing campaigns in the banking sector.
|
Project duration
5 months
Team
- 1 product manager
- 1 designer
- 1 data analytic
Сlient
Lithuanian Bank
TRENDS FOR THE USE OF DATA
SCIENCE TOOLS IN THE BANKING SECTOR
Marketing
- Personalized marketing
- Real-time marketing
- Customer segmentation
- Building social graphs, etc.
Sales
- Predictive analytics
- LTV forecast
- Forecast churn, etc.
Security Service
- Identification of clients making fraudulent transactions
- Detecting hidden connections and "dark" schemes, etc.
Customer Service
- Chatbots to respond to appeals, etc.
Risks
- Short credit questionnaire
- Credit scoring automation, etc.
THE EXPECTED EFFECT OF USING
MACHINE LEARNING ALGORITHMS
Improving the effectiveness of marketing campaigns
MACHINE LEARNING:
Classification
+
Clustering
Classification
+
Clustering
1. Growth of conversions from mailing lists
2. Reducing ineffective contacts
3. A deeper understanding of customers
2. Reducing ineffective contacts
3. A deeper understanding of customers
THE STAGES OF SOLVING THE TASK
OF IMPROVEMENT IN THE EFFICIENCY
OF SENDING TERMS WITH
MACHINE LEARNING ALGORITHMS
THE GENERAL DIAGRAM
CHARACTERISTIC OF THE STAGES.
DEVELOPMENT OF THE PREDICTIVE MODEL
The task
- Using machine learning algorithms determine for all bank customers the probability of applying for a product after receiving an advertising message.
- Form a mailing list of clients with a high probability of applying for the product.
STAGES OF SOLVING THE CLASSIFICATION PROBLEMS
Data collection, data loading into the analytical tool
1
Data preparation for analysis
2
Converting data to numeric format
3
New attributes formation
4
Trait selection
5
Data normalization
6
Development of models classification
7
Quality assessment of models
8
Result output
9
DESCRIPTION OF THE STAGES.
DEVELOPMENT OF THE CLUSTERING MODEL
The task
- Identify homogeneous clusters using machine learning algorithms
- Interpret the results - describe the portraits of the selected clusters
Clients used to build the model
- Clients were used for the cluster analysis,
- probability of response for which >=0,7
Number of dependent variables
- 24 - segmentation was conducted based on data on gender, age, place of residence, as well as on data on spending for three months by spending category
Number of clusters and clustering algorithm
- The mailing list was divided into 5 clusters by the algorithms
- K-means method was used for clustering
THE ACTUAL EFFECT OF USING
MACHINE LEARNING ALGORITHMS
| |
Would you like to improve the effectiveness of your marketing campaigns? |
E-Mail: team@castlemarketing.lt
Phone: +370 6 835 68 31
Phone: +370 6 835 68 31
© 2022 Castle
E-Mail: team@castlemarketing.lt
Phone: +370 6 835 68 31
E-Mail: team@castlemarketing.lt
Phone: +370 6 835 68 31
We use cookies to provide the best site experience.
Ok, don't show again