+91 9319977101 sales@transporthub.in

The Machine Learning Approach for Optimization

by | Jan 21, 2019 | Blog

In this article, I will drill down to a specific problem which affects the logistics teams across the spectrum and at the ground level. The problem which has seen introduction of many techniques including optimization packages flogged by small and big IT companies alike. However, it is rather surprising that new analytical techniques like machine learning approaches have been mostly absent among the options. This article draws from our encouraging experience of deploying the modern techniques with one of our clients. It showcases the efficiency of vehicle loading can be improved by more than 90% and logistics costs saving may touch double digits. The problem I am referring to is of loading the vehicles with optimal load. The problem affects day-to-day operations of almost all the users encompassing consumer goods companies, agriculture, ecommerce companies etc. The loading problems are more acute for the users with varied stock keeping units (SKUs). In other words, heterogeneous mix of cargo poses greater problem than the homogenous loads. For example, food grains in bulk or break bulk are easier to load optimally than the FMCG products. Heterogeneous cargos, such as FMCG or ecommerce, are typically a motley mix of different products, different unit weights, different volume and different packaging. The problems get compounded by the types of available vehicles. The vehicles may be of different size from required sizes to transport cargo optimally. In present scenario, specialized teams jostle with the problem at each plant/depot level. The team’s primary responsibilities include continuous assessment of available load, find suitable vehicles in time and load each vehicle most optimally. The team usually gets short notice on the volume to be dispatched which makes the job even more difficult. Another problem is that expertise gained by experienced personnel is difficult to replicate. If the concerned person leaves, entire system gets into disarray. There is some help available from legacy optimization softwares. However, their recommendations have to be augmented/ verified by human expertise. In fact, many users have complained that the softwares have added to the complexities and increased the work load. The major problem with the available optimization software is that they work on rule-based logic. It is akin to conceiving all the permutation and combination and baking it into rules of the software. The task of thinking ahead all the possibilities is more complex than what most people think. Similar problems have stalled the growth of speech recognition, game programs , translation programs etc. However over the last decade or so, quantum leap has been realized with great advances in new analytical techniques and cheaper computing power. New tools such as Computational Neural Network( CNN) , Support Vector Machines(SVM ) etc. have given newer ways to analyze and categorize the data. We undertook the analysis of varied load mix data from one of leading packers and movers in India. The analysis was carried out with our proprietary algorithm which contained modules for : • Pre-processing the data • Analytic frameworks for categorizing the data utilizing various techniques • Determining the best framework • Module for inputting and prediction of new data It may be noted that even with the limited data availability, the accuracy of prediction was upwards of 75%, which is almost as good as the predictions made by humans. In future, with better data collection and availability of more data, the accuracy is likely to top 90%. Moreover, further tweaking of our proprietary algorithm will also boost accurate predictions. Overall, we are confident that accuracy will be higher that 95% and should top 99% over a short period of time. The algorithm is likely to significantly reduce the logistics costs by loading just the right type, weight and volume of the load. This will ensure the most efficient utilization of space. Moreover, by reducing the efforts by almost 90% , automation through machine learning can free up the resources for other value-added activities. The automation will focus the energies of logistics team to right areas. The teams can concentrate in improving the design of logistics chain, implementing the soft and hard changes which have been hitherto neglected owing to paucity of time etc. In conclusion, it is high time to step-up the pace of adoption of state-of-the-art techniques to make businesses ready for the new word. If the businesses cling to old ways, it may affect the well being of the businesses and jeopardize their chance of surviving in a fast changing world. Contact us : Rohit Chaturvedi partner@kitzoadvisory.com

Headquarter :

543 A-wing,

Balaji Bhavan,Sector 11,

CBD Belapur,Navi Mumbai

Maharashtra 400614, India.

Contact: +91 9319977101

Email: sales@transporthub.in