Data Science North East: September Meeting

Details
Schedule:

6:00 - Food and drinks
6:30 - Bashar Hasan, Head Of Research And Development at Caspian
7:00 - Fernando Torres, Data Analyst at St. Cuthbert’s Hospice

Talk 1- Bashar Hasan - Classification Probability Calibration

Summary:

Classification probability is not intrinsic to most classification approaches. This raises serious challenges in interpreting the output of a classifier in real-life scenarios. Most solutions to assign a probability to the output of a classifier are inadequate and produce either over-confident or under-confident classifiers. Probability calibration the process of transforming the classifier’s probability in a way that better reflect the quality of the model and the interpretability of the results.

Talk 2 - Fernando Torres - Disrupting traditional methods with creativity, a Machine Learning approach in Revenue Management

Abstract:

With a successful start around the 80’s in the Airline industry, Revenue Management was forming what is now called Data Science.

Selling a product at the correct time with the correct price may be a difficult task. However, airline companies and big hotels have been doing it for decades. But traditional methods are complex, as forecasting demand can be challenging, specially for small and medium sized hotels, where the available data can be limited.

With the use of open data, an Expedia API and even the weather, it is possible to increase dimensionality and apply Machine Learning techniques to predict the best hotel’s room price at a given time.

ARIMA modelling for forecasting, Random Forest algorithm and specific R packages are going to be part of a data journey, where all the pieces together can answer a question. What price should I charge?