Worum es bei uns geht
Bevorstehende Events (1)
We are incredibly happy to announce our next Meetup on September 29th at Norcom.
- 2 talks (each ca. 40 min incl. discussion)
- Time for networking + food + drinks before, in between and after the presentations
- Talks are held in English
- We will be taking photos and/or film footage at the event. These will be used to share news about our meetups, and to publicize upcoming events.
The line up:
Markus Böbel - Finding Root Causes in the Sea of Vehicle Data:
Expert Knowledge, Predictive Analysis, and Recommender Systems to the Rescue
The world of vehicle data is massive. From telemetry data and error codes to human written notes on incidences: The heterogeneity is as enormous as the dataset itself. However, in case of reoccurring damages, the data can give hints on the circumstances of the incidence. In this talk we want to present two tools we developed to support our customers with this task. As the first tool, we will have a look at using a recommender system to help the mechanics find parallels between respective error events. For the second tool, we focus on the preprocessing of the data. Here, features such as vehicle profiles, can be created based on expert inputs. The information is fed into an ML-Algorithm to predict the occurrence of damages and errors. At last, we try to derive assumptions on root causes based on the model's feature importance.
Markus Böbel made his Masters in Information Systems at the Technical University Munich. His focus lies on Data Science on time series data. At NorCom he now manages and works on data science projects in the automative domain.
Claudio Giancaterino - Unsupervised Learning applied to the Customer Lifetime Value
The core business of Insurance Companies is to enable individuals and firms to protect themselves against rarely events paying a small premium compared to the eventually damage incurred. Customer Lifetime Value (CLV) evaluates the value of the customer for the Company, in other words, it’s the Net Present Value of the cash flows ascribed to the relationship with a customer. In this work, from the collection of portfolio contracts by one insurance year, will be predicted the Customer Lifetime Value of the last three months of the year, also looking at the effects coming from the use of Unsupervised Learning.
Unsupervised Learning describes tasks that involves using a model to discover a good internal representation of input data useful for subsequent Supervised Learning. In this job Unsupervised Leaning are used to provide a low-dimensional representation of inputs and clustering numerical variables to provide a better portfolio analysis of customers. In both situations, Unsupervised Learning can be used as feature engineering.
Topics: Customer Lifetime Value evaluation with Supervised Learning. Prediction using low-dimensional space representation and prediction with augmented space obtained by dimensionality reduction techniques. Clustering employed as a portfolio analysis tool and clustering used as features engineering.
Claudio Giorgio Giancaterino is an actuary for job and a data science enthusiast in the free time. In recent years, he started loving data science attending Kaggle competitions / hackathons and then
spreading data science in the practical lessons of Insurance Statistics at the Catholic University of Milan, writing articles and attending conferences with workshops / talks.