How to use AI Machine Learning in B2B Marketplace
Machine Learning has certainly come a long way. The confluence of factors such as the evolution of data growth, advancements in computational algorithm and faster machine processing helped create an ideal environment for Deep Neural network and AI to finally gain adoption in the main stream.
Now we have a mother load of data thanks to increasing amount of social sharing and rise of digital devices, Internet of Things (IOT) etc etc.
Just look at the data we have accumulated over the last decade since the creation of Online streaming, Social Media, Mobile and Internet of Things (IOTs).
We are creating about 1.7MB of new information per human being on the planet. Every minutes about 300MB of video is uploaded to Youtube. The accumulated data universe just surpassed 1 zettabyte in 2017. This data usage is expected to triple by 2020. So we are looking at 3 Zettabyte by 2020.
Google processes over 40,000 search queries every second on average which translates to about 3.5 billion searches per day and 1.2 trillion searches per year worldwide.
We are seeing an explosion of connected devices such as Watches (Apple Watch), Home appliances (iRobot Roomba, Control4 Remotes and Touch screens) and Autonomous cars (Tesla), Lidars, Sensors, Surveillance, Home Automation all connected to Internet of Things (IOT), High Resolution images, contributing to the data growth. Not to mention all the data from Facebook, Twitter, Instagram.
Just wait for the 5G network and 4000 satellites
In order to utilize the deep neural network we need a large data set to work with. But we don’t want too much data but just enough to train and optimize the neural network.
What areas to use Deep learning?
There are many areas where you can use deep learning in any organization. But the type of neural network algorithm you use depends on the type of application you want to develop.
- Recurrent Neural Network (RNN) — Any application that utilizes temporal data (one dimensional time series) such as Audio or Speech recognition or machine translation from English to Chinese utilizes RNN model.
- Standard Neural Network (SNN) — A great example of this is Online Advertisement where you want to capture user info when user clicks on ad and track the conversion rate. Another example would be to predict home prices based on certain input variables like zip code, school system, walkability factor etc. Both examples will greatly use SNN model. SNN relies on structured data like a database of data like housing data.
- Convolutional Neural Network (CNN) — This would be something like Image recognition. By tagging images that generates an output 1….x we can match a specific image among a large data set to identify the image you are searching for. Google already does photo search.
- Custom Neural Network — This will be a combination of Convolutional Neural network and SNN where you are utilizing data from multiple sources like on-board cameras, sensors, radar. Examples include Autonomous driving, Robotics control and Maneuvering in Industrial Engineering.
How we plan to use Deep Learning in our startup
I currently work for a startup — InPlaza — which is a marketplace trading platform that connects both exporters and importers from different countries/continents to conduct trade/transactions in an open, transparent, efficient and secure way.
- Exporter/Importer verification — Our exporters and importers are our biggest assets. We strive to bring the best qualified users on our platform. Unlike Alibaba’s platform we plan to do proper vetting and verification of our users before they can start conducting transactions in our platform. We will use inputs from public database (Dian in the case of Colombia), current trade data (import and export data), historical transaction value, credit background, social media presence etc to feed into our Standard Neural network algorithm to help vet and verify users. We continue to train the system with the data we collect, adjust the cost function to optimize the output.
- Product recommendation — Based on buyer’s previously purchased history and browsing history we can train and build the system to power the Recommendation section personalized for the user.
- Logistics supplier matching — Logistics plays an important role in the business transaction. We are using a logistics marketplace system such as FreightOS. The inputs to the deep learning system will be shipping quotes from FreightOS, delivery methods, expected customer delivery date, contract terms, order quantity, country of origin, previous purchase behavior, average order quantity etc.
- Customer Service — We review our users previous interactions, orders, browsing history, compliments and complaints to train the system continuously to server our customer needs in a way it is personalized for them
- Trade Financing — Some of our customers (both exporters and importers) might need to borrow in order to finance their trade. We can build a deep learning algorithm that takes the following inputs: credit and borrowing history, order shipped on time history, customer feedback, trade amount, country of origin etc.
We continue to find opportunities to use Machine learning to automate some of the mundane processes that can derive the price efficiency and fast on boarding of customers on our platform. At the beginning we will not have a ton of data to help train the Deep learning system. But as we gain more traction with transaction volume and customers we will have more data to feed into different algorithms that can help optimize the outcome we desire from the AI system.
“Just as electricity transformed almost everything 100 years ago, today I actually have a hard time thinking of an industry that I don’t think AI (Artificial Intelligence) will transform in the next several years.” ~Andrew Ng, Baidu
Sources:
http://www.internetlivestats.com
