Search

EP-4485301-B1 - METHOD FOR PROFILING EMOTIONAL TYPES

EP4485301B1EP 4485301 B1EP4485301 B1EP 4485301B1EP-4485301-B1

Inventors

  • Cecchetti, Luca
  • Handjaras, Giacomo
  • Roversi, Raffaella
  • Brizio, Matteo
  • Monferrato, Carla
  • Maggi, Francesca

Dates

Publication Date
20260513
Application Date
20240625

Claims (10)

  1. Method for profiling the emotional type of one or more users, comprising the steps of: - providing (1) a database containing a first series of data relative to a plurality of users and a plurality of indicators associated with the first series of data, the first series of data comprising for each user a respective set of data associated with the respective user, each set of data being represented by said indicators; - providing (2) a processing unit in signal communication with the database, said processing unit being configured to analyse and process each set of data of the first series of data; - selecting (3) a sample of users of the plurality of users; - acquiring (4) a second series of data comprising a respective set of qualitative data for each user of the sample of users by means of at least one data entry form, the acquiring step (4) comprising the sub-step of storing (4a) the second series of data in the processing unit; - providing (5) a partitioning algorithm resident in the processing unit; - profiling (6) the users of the sample of users by means of the partitioning algorithm, the profiling step (6) comprising the following sub-steps: - training (61) the partitioning algorithm on the second series of data for each set of qualitative data of each user, - defining (62) an optimal number of clusters, each cluster being related to a different emotional type; - classifying (63) by means of the partitioning algorithm the users of the sample of users by assigning each user to a cluster of the optimal number of clusters on the basis of the respective set of qualitative data of the second series of data; - providing (7) a classification algorithm resident in the processing unit; - profiling (8) the users of the sample of users by means of the classification algorithm by classifying each user into a cluster of the optimal number of clusters on the basis of the indicators representing the respective set of data of the first series of data; - evaluating (9) the classification algorithm as function of a comparison between the profiling of the sample of users obtained by means of the classification algorithm and the profiling of the sample of users obtained by means of the partitioning algorithm, said evaluating step (9) comprising the sub-steps of validating (91) the classification algorithm when there is compatibility in the comparison of said profilings, or repeating (92) the steps of providing (7) a classification algorithm, providing a different classification algorithm, profiling (8) and evaluating (9) the classification algorithm when there is no compatibility in the comparison of said profilings; - profiling (10) each user of the plurality of users by means of the classification algorithm by assigning each user to a cluster of the optimal number of clusters, providing in input to the classification algorithm the indicators representing the respective set of data of the first series of data, to define the emotional type of each user.
  2. Method for profiling according to claim 1, wherein: - the first series of data is related to the banking behaviour of the plurality of users and/or to the money management attitude, each indicator comprising a dichotomous variable representative of a particular banking behaviour and/or of a spending preference; and - the second series of data comprises qualitative data relative to one or more of purchasing and/or investment preferences, financial lifestyle, source of income, aptitude, personality, future prospects, economic self-assessment.
  3. Method for profiling according to claim 1 or 2, wherein: - the step of providing (5) a partitioning algorithm comprises the sub-step of providing a plurality of partitioning algorithms; - the step of profiling (6) the users of the sample of users by means of the partitioning algorithm comprises the sub-steps of: - training (601) each partitioning algorithm on the second series of data for each set of qualitative data of each user; - defining (602) for each partitioning algorithm an optimal number of clusters; - for each partitioning algorithm, classifying (603) the sample of users by means of the respective partitioning algorithm by assigning each user of the sample of users to a cluster of the respective optimal number of clusters based on the respective set of qualitative data; - the evaluating step (9) comprises the sub-steps of: - for each partitioning algorithm, evaluating (901) the classification algorithm as a function of a comparison between the user sample profiling obtained by means of the classification algorithm and the user sample profiling obtained by means of the respective partitioning algorithm; - comparing (902) the partitioning algorithms on the basis of the compatibility towards said profilings; - identifying (903) the optimal partitioning algorithm as a function of the greater compatibility that emerged from the comparison made in the previous sub-step.
  4. Method for profiling according to any one of the preceding claims, wherein the step of acquiring (4) a second series of data comprises the sub-steps of: - providing (40) each user of the sample of users with the data entry form, the data entry form being fillable in by the user; - receiving (41) the data entry form filled in by each user of the sample of users; - storing (42) the data entry form in the processing unit, each data entry form comprising data defining the respective set of qualitative data of the second series of data, the processing unit being configured to analyse and process each set of qualitative data of the second series of data.
  5. Method for profiling according to any one of the preceding claims, comprising the further steps of: - selecting (11) users belonging to one or more clusters of the optimal number of clusters; - transmitting (12) predefined information only to the selected users.
  6. Method for profiling according to any one of the preceding claims, wherein the step of providing (1) a database comprises the sub-step of selecting (101) an optimal number of indicators among the plurality of indicators associated with the first series of data, the optimal number of indicators being comprised between 40 and 60, preferably equal to 50.
  7. Method for profiling according to any one of claims 1 to 6, wherein the partitioning algorithm is of the unsupervised type and/or comprising one of K-means, Spectral Clustering and t distributed Stochastic Neighbour Embedding, preferably of the k-means type.
  8. Method for profiling according to any one of claims 1 to 7, wherein: - the dimensionality of the space of the clusters is defined by a dimension comprised between 1 and 4, preferably equal to 2; and/or - the distance relative to the partitioning algorithm is of the cosine or Euclidean type, preferably Euclidean.
  9. Method for profiling according to any one of the preceding claims, wherein the classification algorithm is of the supervised type and/or comprising one of K-Nearest Neighbors and Support-Vector Machine, preferably K-Nearest Neighbors.
  10. Method for profiling according to any one of the preceding claims, wherein the optimal number of clusters is comprised between 2 and 10, preferably equal to 5.

Description

SCOPE OF APPLICATION The present invention relates to a method for profiling users in clusters related to distinct emotional types. This method finds particular application in customer management, for example in a banking context. Description of the prior art In the state of the art, methods for classifying users on the basis of the personality or behaviours of the individual user are known. In particular, these methods involve the application of multi-dimensional models, i.e. models built according to a multitude of factors related to the common behaviours adopted by the users to be classified. By way of example, known methods provide for the acquisition of data relative to users, for example through self-assessments, based on the perception that users have of themselves, their personality or their behaviours. An example of a method for segmenting customers based on common characteristics in banking is shown in the document "Salman Mousaeirad: Intelligent Vector-based Costumer Segmentation in the Banking Industry, Arxiv.org, Cornell University Library". A further example of a method for customer segmentation is shown in the document "Mahmoud Salaheldin Kasem et. Al.: Costumer Profiling, Segmentation, and Sales Prediction using AI in Direct Marketing, Arxiv.org, Cornell University Library", in which the aforementioned method allows certain customer actions to be predicted in order to increase sales performance. US 2011/270780 A1 instead shows a method for the identification of a user's financial personality and for the construction of a respective risk profile. The publication : "Intelligent Vector-based Customer Segmentation in the Banking Industry", ARXIV.ORG, CORNELL UNIVERSITY LIBRARY, 201 OLIN LIBRARY CORNELL UNIVERSITY ITHACA, NY 14853, 22 December 2020, refers to Customer Segmentation is the process of dividing customers into groups based on common characteristics. The publication "Customer Profiling, Segmentation, and Sales Prediction using AI in Direct Marketing", ARXIV.ORG, CORNELL UNIVERSITY LIBRARY, 201 OLIN LIBRARY CORNELL UNIVERSITY ITHACA, NY 14853, 3 February 2023, refers to the creation of a customer profile and forecast for the sale of goods. Problem of the prior art In the prior art, however, data collected on the basis of users' self-assessments and self-perceptions are of a qualitative nature; therefore, they are difficult to correlate and use in a customer management context. By way of example, in a banking context, it is difficult to correlate the personal data collected relative to the behaviour of a user to the banking dynamics of that user. In addition, in order to classify users into different groups, it is necessary to acquire and process this kind of specific personal data. In fact, the known methods are based on the use of data collected for the individual user, making the acquisition operation and the consequent processing of the data difficult, as well as requiring a high computational cost. Consequently, profiling into emotional types is poorly applicable in the management of a large number of customers, for example, in a banking context. SUMMARY OF THE INVENTION The object of the invention in question is to obtain a method for profiling the emotional type of one or more users in order to be able to overcome the drawbacks of the prior art. The object of the invention in question is to provide a method for profiling the emotional type of one or more users that allows predicting the emotional type of one or more users quickly and efficiently. In particular, the object of the invention in question is to provide a method for profiling the emotional type of one or more users which allows the computational cost and the time required for the profiling of the users to be decreased. The technical task specified, and the purposes specified are substantially achieved by a method for profiling the emotional type of one or more users comprising the steps set forth in one or more of the claims set forth herein. Advantages of the invention By means of one embodiment, it is possible to obtain a method for profiling the emotional type of one or more users that allows users to be classified according to specific indicators relative to quantitative data of the context in which the method is used. In particular, the method involves using an optimized algorithm to determine an optimal number of clusters, representative of the distinct emotional types of users, and classifying users into their respective clusters. Advantageously, the method of the present invention allows user profiling to be extended to the full range of users of the context in which the method is embedded, without the need to acquire personal data from each user. In this way, it is possible to save time and resources and decrease the required computational cost. Still advantageously, the method of the present invention allows the experience of the individual user to be customized. In fact, the method allows the selection of a s