We are Solution oriented
Solution-oriented The goal of Agile Data Decisions is to create value from your existing data assets by providing technology, innovative software and consultancy services. We would like to help you make your existing digital data relevant to, and useable by, your decision platform.
iQC is a machine learning system developed by Agile DD for the Oil & Gas industry. iQC encapsulates the key philosophy of Agile DD:
Machine learning systems must learn rapidly from human feedback,
Expertise is with the end-user and not with the machine,
Link between information and source of information is key to monitor the quality of datasets.
iQC will extract for you the validated well metadata you need
We consider our customers to be the experts, they have the domain skills necessary to drive the building of machine learning models and they have the experience to know which piece of information is more relevant than another.
Nevertheless, we can provide backup for your team to inject our experience in accessing unstructured data. This will ensure a rapid and effective uptake and the best results.
AgileDD presents "Auto-Predict", a machine learning library to forecast time-series. Combined with iQC, Auto-Predict can be used in conjunction with extracted information from unstructured documents. Historic information from legacy unstructured documents and structured data can then be combined in Auto-Predict to predict quantities such as supply or demand, inventory, or for asset monitoring and failure prediction.
iQC is the first machine learning system developed by AgileDD for the Oil & Gas industry.
iQC embodies the AgileDD principles:
Machine learning systems must learn rapidly from human feedback;
Expertise is with the end-user and not with the machine;
The link between information and the source of that information is key to monitor the quality of a dataset.
iQC will extract for you the validated well metadata you need from your unstructured datasets (cataloguing) and will classify your documents according your taxonomy (indexing).
Developped using a Big Data IT environment and dedicated to unstructured information, iQC has no limit in terms of number of documents, number of wells, amount of metadata, file formats, file sizes
OCR Optical Character Recognition
Since we index the files according their contents and we extract text metadata, our process includes an OCR step for non text searchable documents.
But the layout of documents may be quite complex, therefore we apply a specific pre and post OCR processing to make possible an efficient pattern recognition on the text and associated features.
Once OCR is applied, iQC classifies the documents according their contents and using a taxonomy provided by the user.
The documents are grouped per extracted entities (e.g. names of oil wells) and category. The classification confidence for the entity and category is displayed as a colored square. Confident classifications are green, not-confident classifications are red.
The stack power
iQC extracts not only the metadata listed by the user from each of the documents but also groups the extracted metadata, for example, by well name or other entities. Knowing that iQC is able to score an extraction according to its confidence, we can propose to the expert (you!) the best candidate for a metadata value and its associated confidence. We can display the histogram of all candidate values to illustrate the variation of the same metadata among all the documents of the same type (or having the same entity such as a well name).
QC the results and train the machine in one step
You know the type of documents you are manipulating, you know the data you are targeting, you are the EXPERT, therefore we believe you are the best trainer to train the machine. That's why we invested so much in the graphic user interface: to make the metadata QC and the machine training an enjoyable experience.
Any extracted metadata is selectable, searchable in the documents and accurately located in them. If you fix an incorrect result, or in the opposite, you validate a positive one, iQC will remember and be sure the the metadata extraction will be better next time!
You want to know more? Have a look on The Leading Edge paper about iQC (March 2017):
Move iQC to the data or the data to iQC ?
iQC is flexible enough for you to upload your documents on the cloud and make your metadata extractions on a easily scalable environment, or in the opposite to implement it locally, on your private network, close to your data.
Both are possible ... and more!
At Agile Data Decisions, we believe that the cloud is the best place for most data management tasks because of the scalability of the cloud resources which can be adjusted to your computing needs.
Agile DD has selected Microsoft Azure as the cloud provider to guarantee to our customers an efficient and secure user experience.
Azure advantages have been highlighted recently by Gardner in their 2016 Quadrant for aPaaS. See here.
Using iQC in the cloud, you will take advantage of our existing learning models and will participate in their improvement
Your documents are very sensitive and numerous, moving to the cloud may not make sense. In addition, you aren't interested by the collaborative experience for improving the learning models: Implementing iQC locally on your private network is probably the best solution in this case.
This purely local implementation of our applications will obviously not stop you from benefiting from Agile DD's support: new versions are available for download on the web.
We can enable a fully-managed Agile DD experience in your own environment.
You want to use the most updated learning models available on the Agile DD cloud but you don't want to move your data to the cloud. By the way, moving millions of legacy unstructured files on to the cloud may be a long and difficult process.
In this case, we have designed a Hybrid version of iQC which accesses the files of your local network but run the Machine Learning tasks on the AgileDD Azure cloud where the most advanced learning models are available in a collaborative mode for all cloud users.