DataTalksBusiness

In the bygone era, in the lush villages, our Nanny and her neighbours, in their daily evening meet would mostly talk about something new that happened in their village. Sheela aunty’s fresh pickle would have brought “Wooos” or that look from  Anju’s aunty that said “Yumm! I want more!” or the wafting smell would have made Geetha aunty come running to her house. In the end, it’s a success story for the freshly made pickle. Analytics and data science is all about aping the above scenario – Capture people’s senses, quantify and qualify them to extract meaningful insights. In this process the five senses of humans – visual, auditory, smell, taste and touch which are processed by human brains to validate a scenario, is now being captured in the virtual world through social media platforms. If reviews on Facebook/Zomato provide verbal feedback about a new hotel, video recordings along with audio commentary on live service of the delicacies provide visual and auditory data. These ar...

Demand forecasting is like the pivot point in a see-saw, with suppliers on one side and the customers on the other; warehouses, production units, wholesalers and retailers make up the rest of the chain, all of them being the reservoirs of different forms of inventories. Hence a miscalculation of demand in any place of this chain would cause an imbalance in the see saw, affecting the manufacturing time, delivery time, pricing, the brand name and most important of all, the trusted customers of the end product! An underestimation of inventory would result in the baker giving you a cake without the cherry on the top. And an overestimate would result in him giving you one cake free! And that’s when the devil in us would say “Maybe he’s trying to give away stale cakes!”. Excess stock would lead to bloated inventory and high cost incurred by the company. Underestimating demand means you are losing your valued customer to your competitor. Absence of any forecasting would lead to the bul...

To opine on the given statement, I would say that “It is  not always mandatory  to do the classification exercise to predict the target variable.” To elucidate, I would like to use the supervised learning methods- Classification and Regression with an example of the  Hurricane Matthew . Hurricane Matthew which hit the state of Florida was  categorised as 4  amongst the categories 1,2,3,4 and 5. This  classification  into a predefined class would be done using classification algorithm which uses input variables like the wind speed, ocean currents, temperature, terrain etc. For evacuation purposes, values like the  maximum speed  of the hurricane winds or the maximum number of days the winds prevails in an area, would have been computed using  regression  techniques. This does not necessitate the output variable to be classified into groups. Classification technique, predicts if a certain observation belongs to a predefined cl...

The advent of AI in retail and CPG sector has revolutionised the way shoppers buy and the way buyers sell. The new cult is knowing your customers and differentiating their needs. If "Personalisation" is one story, taking the current global trend to the customers door step is another. From warehouses, to design studios, to your eCommerce websites and finally to delivery of the goods, AI is playing a huge role in the automating process. AI is being used to address problems such as speech recognition, natural language understanding, question answering, dialogue systems, product recommendations,product search, forecasting future product demand, etc.  Bengaluru-based Locus is enabling logistics for companies like Quikr, Delhivery, Lenskart and Urban Ladder using supply chain analytics. Another startup, www.stylumia.com , applies AI and computer vision, to analyse the world of fashion and predict the latest trends in the fashion industy. Myntra's AI projects is to come up...

Just like the choice of placement of a bushy Bonsai shrub or a tall coconut tree in our garden depends on various parameters, so is the decision on the type of CHAID tree we want to construct. A number of factors like the type of variable, the number of variables, the categorisation level of each variable, the depth of analysis one wants to conduct, size of data set, etc, influences the choice of the merging and splitting criteria. The higher the alpha2, the denser is our CHAID tree; higher the alpha4, the taller is our tree. The denser tree helps us analyse the impact of various categories of variables on target variables with microscopic accuracy. A taller tree portrays good picture of how the individual variables impact the target variable.   Number of variables Less More Number of  categories Less alpha2 and  alpha4 = 0.05 alpha2 = 0.05, alpha4 = 0.02 More alpha2 = 0.02 alph4 = 0.05 alpha2 and  alpha4 = 0.02

Being from an IT background, the first thing that flashed me while reading about post-optimality analysis/ sensitivity analysis was the Disaster Recovery activities the management trains its resources to be equipped with. (Equip a pre-existing team handle uncertainties). The optimal solution found for various scenarios is always the best for the that scenario. But in life nothing is constant and predictable. So doing a post optimality analysis enables to better equip the solution thus suggested. It not only iterates the boundary limit within which our solution works best in but also encompasses the maximum amount  fluctuation our solution can handle.