Neosperience Machine Learning Applications

Focus on machine learning applications enabled by Neosperience Cloud adoption

Transcript

1. www.neosperience.com | blog.neosperience.com | info@neosperience.com October 2018 Neosperience. Empathy in Technology focus on Machine Learning
2. A model for time series analysis built in collaboration with fin-tech players and research institutes Analyzing Time Series In an effort to build click/scroll stream analysis tools, Neosperience started working with finance sector and University of Pavia to tune existing State of the Art models in order to guarantee high throughput and accuracy. Streams of data are analyzed on a GPU Rig applying LSTM neural networks and building a state of the art model to be used as a basis for model evolution. After a timeframe of 12 months of work of a joint team of data science experts, this joint effort produced significant improvements in this models making Neosperience able to crunch, both on premise and in cloud, huge streams of data to build a realtime model capable of: • temporal abstractions over data streams • patterns of data behaviour • clusters of patterns • cluster classification Neosperience model can be leveraged to implement effectively: • anomaly prediction • event prediction • feature clustering and engineering
3. Implementation of State of the Art CNN models within Neosperience Cloud Image/Video classification and recognition models Collaboration between Neosperience and retail brands, led to development of a novel model for image classification, from a decade of scientific works on neural networks. Convolutionary Neural Networks represent the state of the art of image classification, but still miss optimization to specific use cases: they can be applied to a wide range of analysis, but fine tuning and hyperparameter optimization still have to be addressed. Neosperience worked since 2016 on model improvements with Computer Vision Lab of University of Pavia
4. Customer Experience and Channel Optimization • Discover the optimal way to reach a customer with certain characteristics • Omnichannel service delivery optimization • Conversational interfaces (chatbots) • Customization of communication, i.e. “One Customer” initiative
5. Customer Segmentation • Allowing the Bank to understand qualitatively its different customer groups, enables to deliver them tailored experiences, plus answer questions like: what makes customers buy, stop buying, etc. • Product mix; what mix of products offers the lowest churn? eg. Giving a combined policy discount for life + auto = low churn • Discount targeting; what is the probability of inducing the desired behavior with a discount
6. Lifetime Value (LTV) • Predict the characteristics of high LTV customers; this supports customer segmentation, identifies upsell opportunities and supports marketing initiatives • Wallet share estimation; calculating the proportion of a customer's spend in a category to identify upsell and cross-sell opportunities • Target market; understanding the target helps determine exactly what your products or services will be, and what kind of customer service tactics work best • Lead prioritization; what is a given lead's likelihood of closing
7. Upselling and Cross-selling • Given a customer's past browsing history, purchase history and other characteristics, what are they likely to want to purchase in the future? • Identifying commonalities among customers for a specific product, we are able to discover the ideal prospective customers to upgrade that product
8. Churn • Working out the characteristics of churners allows to produce adjustments and an online algorithm allows them to reach out to churners; show the characteristics of low wallet share customers • Reactivation likelihood; what is the reactivation likelihood for a given customer; how can we reactivate it
9. Customer Personality (1/2) • Identify risk-taking attitudes (financial risk propension vs. aversion) • Personality traits that influence investment decisions (e.g. Need for Cognitive Closure, openness to new information, and tendency to update vs. not to update one’s investment portfolios in a way that reflects one’s risk preference) • Personality traits that influence risky investment decisions (e.g. Locus of control, optimism, and willingness to invest in risky assets)
10. Customer Personality (2/2) • All possible traits that determine the most effective communication style to adopt with a user. These are 19 Traits: Big 5, Locus of Control, Need for Cognitive Closure, Need for Cognition, Need for Affect, Impulsiveness, Self- efficacy, Sensation Seeking, Optimism, Need to Belong, Narcissism, Need for Uniqueness, Sense of Power. • Personality traits that allow to better determine user needs, helping to improve online payment services utilization and to attract customers who would not normally consider online credit card payment services. • Personality traits that help to predict money attitudes (e.g. saving, spending, impulse buying, etc.).
11. Trading • Trading automation • Predictive modelling e.g. of securities Financial Advisory • Cross/up sell investment services based on demographics/past investments/portfolio composition • Personalized recommendations (loans, credit offers, paid services, ...) • Portfolio optimization & rebalancing
12. Processes Support • Loan / Insurance underwriting • Processes streamlining (handle simple and ‘low-value processes’ with AI) • Document analysis and retrieval, relevant for a process Risk Management, Creditworthiness and Fraud Detection • Credit analysis. Will a borrower repay a loan? • Fraud detection • Terrorist related accounts / transactions • Money laundering transactions
13. Speech-to-text and NLP to empower support Chatbot for support and interrogation • Query internal systems for status and issues • Provide proactive report information • Customer advisor and support • First level customer support
14. www.neosperience.com | blog.neosperience.com | info@neosperience.com !14