How Operations Research Helped Me Decide Which Laptop to Buy

As many of you know, I am a student in the MBA program in Operations Research and Decision Making (Pesquisa Operacional e Tomada de Decisão) at BsBr, coordinated by Professor Dr. Marcos Santos, one of the most amazing people in operations research in Brazil. In this MBA, at the end of each course, students are asked to carry out, in groups, a small practical case with real data using the methods presented. So, precisely in the course on outranking methods, I decided to combine business with pleasure: I merged the assignment with my need to buy a new high-performance laptop for my computer vision work and suggested this topic to my group.

So, together with my group, I ran a search following a few premises:

• High-performance laptops with an i7 processor;
• Laptops from Brazilian retailers, from the brands Asus, Avell and Dell, with at least 1 year of warranty support;
• A slot for an additional hard drive;
• Price for purchase without discounts.

We collected data on the laptops for the following criteria:

• RAM (GB)
• Storage (TB)
• Graphics card (GB)
• Processor (Hz)
• Price (R$)

This gave us the following decision matrix.

To solve the problem, I used the pyDecision Python library and chose the ELECTRE III method, since the goal was to rank the best laptops based on pseudo-criteria, taking into account the weights established for each criterion. This method also has the advantage that the concordance and discordance relations are established by the decision maker (the final buyer, in our case), and it also reveals how strict those relations are.

The preference (p), indifference (q) and veto (v) thresholds, as well as the weights (w), were defined through discussions with the end user, so that these parameters would be as realistic as possible with respect to their actual interests.

When running the electre_iii function, we can choose whether or not to generate the decision graph. It shows which alternatives stood out, and depending on the multi-criteria model you choose, you will get a better answer that can support your decision making.

global_concordance, credibility, rank_D, rank_A, rank_N, rank_P = electre_iii(ds_list, P = P, Q = Q, V = V, W = W, graph = True)

Since the model returns its result as a somewhat clumsy list, I wrote the code below to split it and turn the result into a data frame.

rank_D_final = []
rank = []
k = 1
for i in range(0, len(rank_D)):
    splited = list(rank_D[i].split('; '))
    for s in splited:
        rank_D_final.append(s)
        rank.append(k)
        k += 1
        resultado = pd.DataFrame()        

 resultado["Alternativas"] = rank_D_final
 resultado["Resultado"] = rank

Then I retrieved the actual names of the alternatives from our initial dataset, showing the model's result alongside the name of each laptop.

dic = ds.iloc[:, 0:2].set_index("Alternativas")
dic.join(resultado.set_index("Alternativas"),on="Alternativas").sort_values("Resultado")

We can see that the AVELL A65 MOB is the most expensive of all the laptops I included in the search; however, its graphics card is the best, which is why it stands out. The other laptop tied in first place, the AVELL A70 MOB, is cheaper than the A65 but outperforms the others on the technical requirements, making it the one I am most likely to buy.

With this simple multi-criteria application, I was able to settle my indecision about buying the new machine. Unfortunately, I haven't bought it yet, but I am planning to do so soon, and then I will post a review of it on the Spatial Hackers channel.

Just between us, I am happy that AVELL stood out from the others, especially because it is a Brazilian manufacturer with which I have had a great experience for more than 5 years.

I'll leave this last bit here to thank the brilliant people in my MBA group: Daniella Tombi, Geraldo Jorge, Leonardo Ciotti, Lucas Tayrone and Robsion Gomes.