In industries where equipment is used, and its availability is essential, the capability to forecast the possible failure before it occurs can be the key to saving time and money. The conventional ways of measuring equipment reliability involve using anomaly detection or control charts that focus on changes in some data point or value. However, these approaches may not capture small changes in the behavior of the machinery that may be precursors to failure. This is where Change Point Analysis (CPA) comes into play, a technique that helps to identify changes in the trends of the data rather than the values. By unearthing such latent relationships, CPA provides a superior and preventive means of predicting equipment failure, thus enabling organizations to avoid equipment breakdowns and other losses, thereby increasing equipment durability and decreasing maintenance expenses. When used together with analytics, this method offers a strong instrument for industries that seek to improve reliability and minimize disruptions’ costs.
Ankit Bansal has established himself as a consultant and a theorist in the field of predictive analytics especially in network management. The consultants with such experiences and accomplishments were given onshore lead roles in his consulting company, which proves that his work and leadership were exceptional in facilitating innovation. He considers one of the greatest accomplishments in his career to have been offered a verbal offer by a client, a networking giant, to join the company in a consulting role after doing a great job. Not only did he successfully deliver high-stakes projects that gave him further opportunities, but he was also promoted to the onshore lead which proved that his organization had a lot of trust in him.
Among Ankit’s most effective activities were the specific outcomes achieved for a network client. After completing a large-scale project using predictive analytics for equipment maintenance, he was able to help the company achieve and exceed Service Level Agreements (SLAs), ensuring that network systems stayed operational without being affected by faulty equipment. By using a predictive model to identify which pieces of equipment were likely to fail, the company could send service engineers to fix or replace faulty hardware before it disrupted network infrastructure. "The results were impressive, as SLA quality was enhanced, and the client saved several million dollars each year by addressing issues before they reached critical failure points," Bansal reported.
This implementation was successful in enhancing the efficiency of the client, and at the same time, contributed to the reputation of the client. The improvement of their services enabled the company to acquire between three to four clients because of their better network reliability. It was not just about the predictive maintenance; in the organization, he was able to affect cost and performance in other ways too.
For the second phase of the project, he was given more critical tasks since he had successfully completed the first project, and his expertise was now being used in Phase 2 which entailed applying the results of the predictive model in a prescriptive analytics framework. He also contributed to developing Power BI dashboards to present and manage the results of the model’s performance. Furthermore, Ankit was more actively involved in transforming the model from theoretical research to an actual production model to be implemented in the organization’s operations.
Moreover, among the biggest challenges that he faced, there was one that was connected with data patterns which did not reveal drastic shifts even if rather traditional approaches such as control charts and anomaly detection were used. This led him to look for other sophisticated algorithms that could help in the analysis of the data and patterns that were not initially recognizable, and this was a key factor in the success of the project. Introducing new variables without distorting the results while improving forecast accuracy also proved challenging; each variable had to be rigorously tested to understand its impact on failure probabilities.
Another problem was the complexity of the machine learning algorithms Ankit used, challenges were also observed when applying the new at the time models like gradient boosting and neural networks, which are often referred to as “black boxes”. There were also some challenges in trying to explain how each input variable influenced the probability of failure. However, Bansal was able to overcome these problems by getting special permission to use SAS Enterprise Miner, which has costly licenses, and by concentrating on the optimization of the ratio of false positive results to false negatives. This particular decision to focus on false positives made his strategic decision to replace equipment even if there was only a small probability of failure, paramount to maintaining the client’s network uptime.
In addition to the previously mentioned work, he has also written extensively on methodologies adopted by various organizations. His paper, Introduction And Application Of Change Point Analysis In Analytics Space in the Analytics Space, has been widely recognized for its in-depth exploration of secure API design in cloud-based systems, further solidifying his reputation in the field.
Looking at the future, Bansal has faith that the techniques that he has contributed have their uses in almost every sector of business and technology, not just the hardware and networking fields. According to him, these predictive models can be applied to test data quality in different fields. In contrast to the typical control charts or other methods for detecting anomalies, which are based on value shifts, his approach is based on trend shifts, which are a much more accurate indicator of the problem. This type of analysis based on trends is very useful for identifying sharp fluctuations that may be missed in other cases, so it is an effective tool for optimizing business processes and managing risks.
This content is produced by Rahul Sharma.