“In God we trust; all others bring data.”
-W. Edwards Deming
“We are drowning in information and starving for knowledge.”
– Rutherford D. Roger
The science behind using data, to answer a question can be termed as Data Science. Analyzing data is an art. The analysis starts with a question aided by the data. A question posed for a business need but with inappropriate data will not lead to helpful insights.
At InTimeTec, we recognize and strive to avoid the ‘hype’ around Data Analytics and focus on delivering ‘valuable’ insights to our customers, starting with the questions.
An insightful answer to a specific question based on the data involves an overlap of various disciplines. InTimeTec has been investing in initiatives that combine the synergies of the varied expertise of its engineers across domains.
Our experience and expertise in the fields of Embedded Systems, DevOps, Web Application Development, Database Management combined with our approach to cross-fertilization of ideas, uniquely sets us apart.
InTimeTec has been devoting time and resources to leverage its engineering strengths and reinforce it with math and statistics skills. We are therefore, in a strong position to provide Data Analytics solutions to our customers.
A data analysis effort is fruitful, if it is clear at an early stage as to what type of question is being answered by the analysis. Are we predicting or forecasting? Are we trying to draw inferences from the data or are we seeking information from the data?
Data has to be combed to check for data entry errors, missing data and outliers. Based on necessity, the data may need to be transformed. It has to be explored graphically and statistically to infer relationships and patterns within the data.
Models try to approximate the function from which the observed data is generated. Based on the type of problem, different approaches to model the data distribution can be attempted.
A model has to be optimized so that it is able to fit the data without running the risk of being under-fit or over-fit. The model should be able to generalize equally well for unforeseen data. It is very critical that a model is able to balance various contradicting factors.
Finally, the whole analysis effort should result in convincingly answering the questions about the data.
An instance of supervised learning. Office Imaging and output needs of a customer are collected. The characteristic features of Imaging hardware, along with customer’s requirements are combined to create a dataset.
A model is built that is able to map the requirements to a specific device or device family.
The model can be extended to recommend a new device to replace an older one.
Anomaly Detection Engine
Anomalies are rare events, but can be a major problem when they occur. Given a pattern of data, the engine
should determine if there is any anomaly.
Anomalous patterns can be flagged for immediate investigation. It can benefit the customer by proactively resolving the issues than doing it reactively.
An engine that will provide information on the performance of a device. It’s a reporting module that will provide information using graphical charts and figures.
Compare the performance of a device against a set of similar devices or device family.
Check whether the differences are statistically significant. Graphics and charts to perform a statistical analysis on a pool of machines.