By – Dr. Raushan Singh Assistant Professor, Department of CSE, SRM University -AP ( Amaravati)
Students are skipping the hardest and most important step in research, and their results are paying the price.
Over the years I spent in academia training faculty members and researchers, my work took me across the globe and to most of the cities of India. In classroom after classroom, one trend kept showing up: researchers hesitate to collect their own data.
Take a simple experiment that needs basic atmospheric readings such as temperature, humidity, air pressure and altitude, all of which can be gathered with sensors. Or go a step further to proper weather data that can be used to map storms, cyclones, smog and fog. In most cases, students do not build anything to measure these conditions themselves. They take the easy route and head to a website that offers a ready-made dataset.
This creates many problems. Very often the data comes from another state or even another country, recorded under different conditions and in a different environment. A model trained on such a dataset is already at a disadvantage when it comes to real-world implementation. There can be many arguments about data validation, calibration, cleaning and filtration. But think of a dress sewn properly for one person’s body. It will not fit everyone. Others will have to alter it to suit themselves, and even after the alteration, it will never give the comfort of a garment made for them in the first place. Borrowed data works the same way. If you want to do correct and accurate research, collecting your own data is a must. Data collection is considered the most important and the most challenging part of research. Your outputs are directly proportional to the quality of your data.
This is where the challenges begin. Most of the researchers skip the extra step before data collection to build proper sensors suited to their specific needs. Purchasing components means cost. Then comes integrating them with a microcontroller, setting up communication, arranging a storage unit, and so on. Faced with all this, they go straight for the easier task: collecting data that someone else gathered, which comes with its own flaws.
From my experience, the components needed in most cases are not extremely costly. Most researchers can approach their university or their mentors for guidance on building and procuring data collection units. Doing so allows them to say with pride, “From the first sensor to the final result, this research is mine.“
And that is what leads to correct research practices.
