YouTube2Text

یادگیری تحلیل داده از صفر با AI | بدون این ۶ قدم، Data Analyst نمیشی! — Transcript

by Hozhi · 2,853 words · 471 segments · language en · Watch on YouTube

Full transcript

  1. 0:01Why, now that AI can write SQL, do you
  2. 0:04see companies like Digikala and Snapp
  3. 0:06looking for this very skill when you
  4. 0:08open data analyst job postings?
  5. 0:16Well, why do companies with full access
  6. 0:18to AI still look for people who can
  7. 0:20write queries and know SQL?
  8. 0:24In these same ads, what else are they
  9. 0:26asking for alongside SQL that AI lacks,
  10. 0:28and where should you, with no experience
  11. 0:31, start so that one of these job ads
  12. 0:33actually responds to you?
  13. 0:35By the end of this video, you will have
  14. 0:37a complete roadmap; from the very first
  15. 0:39skill to your first interview. And it's
  16. 0:42all based on what companies in Iran have
  17. 0:44actually written in their job
  18. 0:45descriptions. Even if you haven't
  19. 0:47written a single line of code until now
  20. 0:49, it doesn't matter at all; because we
  21. 0:50are going to go forward step by step
  22. 0:52together.
  23. 0:53For those seeing my video for the first
  24. 0:55time, I’m, and I teach you
  25. 0:57everything from the world of AI that
  26. 0:59impacts my own life in simple terms.
  27. 1:02Well, this path we are going to take
  28. 1:04together has 6 steps. Step one: Excel,
  29. 1:07getting familiar with data without even
  30. 1:09a single line of code. Step two: SQL,
  31. 1:12the most important skill for this job.
  32. 1:15Step three: Basic statistics, so you
  33. 1:17understand if a number is truly
  34. 1:19meaningful or not.
  35. 1:21Step four: Dashboards or Power BI.
  36. 1:25Step five: Python for when data gets big
  37. 1:28.
  38. 1:30Step six: AI, which makes you faster
  39. 1:32rather than replacing you.
  40. 1:35After that, I'll tell you about a skill
  41. 1:37more important than all of these, and
  42. 1:39finally, project building and finding a
  43. 1:41job. For every step, I will tell you
  44. 1:43three things. First, why is it necessary
  45. 1:46?
  46. 1:47Second, exactly what to learn and in
  47. 1:49what order.
  48. 1:51Third, when you are ready for the next
  49. 1:53step.
  50. 1:54Let me tell you one point before
  51. 1:55starting.
  52. 1:56From day one, practice on real data.
  53. 1:59Not fake examples.
  54. 2:01There is a site called Kaggle that is
  55. 2:03full of real, free data.
  56. 2:06It has data from a Brazilian online
  57. 2:08store with about 100,000 real orders;
  58. 2:10what did they buy? How much did they buy
  59. 2:13, when did it arrive, and what rating
  60. 2:15did the customer give?
  61. 2:17All the exercises in this tutorial are
  62. 2:19based on this file, and I will put the
  63. 2:21link in the description of this video
  64. 2:23for you to use.
  65. 2:25Step one: Excel. Why Excel first?
  66. 2:28Because it’s the simplest way to get
  67. 2:30familiar with data. It doesn’t require
  68. 2:32code. Almost every company works with it
  69. 2:34.
  70. 2:35Digikala has asked for Excel and SQL
  71. 2:37together for one of its data analyst
  72. 2:39roles. Exactly. Now, what should you
  73. 2:41learn? In the exact order I tell you.
  74. 2:44First, sorting and filtering.
  75. 2:46Sort and filter.
  76. 2:48For example, being able to see only the
  77. 2:50orders from a single month.
  78. 2:54Next, some basic formulas like sum,
  79. 2:56average, and conditional logic, plus
  80. 2:59joining two tables via a common column (
  81. 3:01the latter is called VLOOKUP); then data
  82. 3:04cleaning, which means finding and fixing
  83. 3:07duplicate rows or empty cells; finally,
  84. 3:09a summary table that aggregates
  85. 3:12thousands of rows with a few clicks,
  86. 3:14known as a Pivot Table, along with a
  87. 3:16chart.
  88. 3:19Let me mention a point specific to
  89. 3:21Persian data. When working in an Iranian
  90. 3:24company, the date is usually Persian (
  91. 3:26Shamsi). In some places, numbers are
  92. 3:28typed in Persian, while in others, they
  93. 3:31are in English.
  94. 3:34The name of a city, like Tabriz, might
  95. 3:37be entered with an Arabic 'ye' or a
  96. 3:39Persian one; the computer considers
  97. 3:41these as two different cities.
  98. 3:45So, we must take data cleaning very
  99. 3:48seriously. If the data is dirty, any
  100. 3:50analysis you perform afterwards will
  101. 3:52likely turn out to be wrong. Now, when
  102. 3:55should you move on to the next step?
  103. 3:58When you can open the orders file in
  104. 4:00Excel, use a Pivot Table to show how
  105. 4:02many orders were recorded each month,
  106. 4:04and have a chart next to it.
  107. 4:07Step two: SQL. Why SQL? It is exactly
  108. 4:10what we saw at the beginning of the
  109. 4:12video in job ads for Digikala, SnappFood
  110. 4:15, Snapp, Basalam, Alibaba—it's mentioned
  111. 4:17almost everywhere. First, let me tell
  112. 4:19you what it is. A company's data is not
  113. 4:22in Excel.
  114. 4:23It is stored in a place where several
  115. 4:25tables are kept together.
  116. 4:28A table for orders, for instance, a
  117. 4:30table for customers, a table for
  118. 4:32products—like several accounting ledgers
  119. 4:34placed side by side.
  120. 4:37This collection is called a database.
  121. 4:39SQL is the language you use to ask
  122. 4:41questions of these ledgers. Each
  123. 4:44question is called a query. Now, what
  124. 4:46exactly do you need to learn?
  125. 4:49In this specific order: First, selecting
  126. 4:52and filtering (SELECT and WHERE), like "
  127. 4:54give me last month's orders"; then
  128. 4:56aggregating and grouping (SUM, GROUP BY
  129. 4:59), for example, "give me the total sales
  130. 5:01broken down by product category"; then
  131. 5:04joining tables together, connecting
  132. 5:06orders to products to see what each
  133. 5:08order contained; this is called a JOIN,
  134. 5:11which is a term specifically mentioned
  135. 5:13in job ads.
  136. 5:17Finally, once you are comfortable
  137. 5:19writing these, move on to nested queries
  138. 5:21and window functions.
  139. 5:24For practice, sites like LeetCode and
  140. 5:26Stratascratch are great resources that
  141. 5:28offer free SQL questions.
  142. 5:31Now, when should we move on to the next
  143. 5:33step?
  144. 5:34When you can answer a question using a
  145. 5:36single query on that Kaggle data, for
  146. 5:38example, "which product category had the
  147. 5:40highest sales?"
  148. 5:42; this question requires all three
  149. 5:44skills together. Before we continue, my
  150. 5:46friends, I’d like to ask that if this
  151. 5:48tutorial is useful to you, please like
  152. 5:51the video, share it with your friends,
  153. 5:53subscribe to my channel, and hit the
  154. 5:55bell icon so you don't miss future
  155. 5:57tutorials; because I create tutorials
  156. 5:59every week that can have a positive
  157. 6:01impact on your life. Step three: basic
  158. 6:03statistics. Now, why statistics? Because
  159. 6:06up to this point you’ve learned to
  160. 6:08extract numbers, now you need to
  161. 6:09understand if those numbers are actually
  162. 6:11meaningful or not.
  163. 6:12For example, Basalam specifically asks
  164. 6:15for the cause in the job posting from
  165. 6:17anyone who wants to fill out the form.
  166. 6:21You don't need to be a statistician, but
  167. 6:23you must know these three; these three
  168. 6:26are enough: first, average or mean, and
  169. 6:28median.
  170. 6:30I’ll give an example so you understand:
  171. 6:32imagine a 10-person company where the
  172. 6:34manager makes 10 times more than the
  173. 6:36others. The average salary comes out
  174. 6:38high, but it doesn't represent the
  175. 6:40salary of most people.
  176. 6:42The median is the middle number when you
  177. 6:44sort everyone from low to high; here,
  178. 6:46the median is more realistic.
  179. 6:50Second, dispersion or variability.
  180. 6:54This means how much the numbers go up
  181. 6:56and down.
  182. 6:58If sales naturally fluctuate by 10 each
  183. 7:00month, a drop of, say, 5 isn't something
  184. 7:03strange.
  185. 7:06Third, correlation, which means two
  186. 7:08things go up and down together.
  187. 7:12A common mistake happens here; for
  188. 7:14example, in the summer, ice cream sales
  189. 7:17go up, and air conditioner sales go up
  190. 7:19too.
  191. 7:20These two have a correlation, but ice
  192. 7:22cream doesn't cause someone to buy an AC
  193. 7:25; both are because of the hot weather.
  194. 7:27This is called correlation, not
  195. 7:29causation (correlation is not causation
  196. 7:32).
  197. 7:33Finally, you should also know one more
  198. 7:35term called hypothesis testing; it means
  199. 7:38checking if a change is larger than the
  200. 7:40normal fluctuation or not.
  201. 7:43You don't need to memorize the formulas
  202. 7:45at first, you just need to know what it
  203. 7:47is at all. When should you go to the
  204. 7:49next step?
  205. 7:51When you can calculate the average and
  206. 7:53median of order amounts in that same
  207. 7:54data and explain why they are different
  208. 7:56from each other.
  209. 7:58Dashboard with Power BI; why a dashboard
  210. 8:01? Because managers don't have time to
  211. 8:03read tables.
  212. 8:04A dashboard is a single page that shows
  213. 8:07important business numbers in one place
  214. 8:09, live; like the speedometer in a car
  215. 8:11where speed and fuel are always in front
  216. 8:13of our eyes. What is the tool for it?
  217. 8:16In job postings, the name you usually
  218. 8:19see most often is Power BI; Snapp, Asan
  219. 8:22Pardakht, and Alibaba all use such tools
  220. 8:25.
  221. 8:28Power BI, but Tableau and Metabase do
  222. 8:30the same thing (you can build dashboards
  223. 8:32with them), but they are used less.
  224. 8:36Now, what do you need to learn for this
  225. 8:38?
  226. 8:41Follow these steps in this exact order:
  227. 8:44first install, then import your data
  228. 8:46file (the desktop version of Power BI is
  229. 8:49free), create a basic chart—like sales
  230. 8:51by month or sales by category—and
  231. 8:53finally, add a filter to the page so the
  232. 8:56manager can choose what to see; this
  233. 8:58filter is called a slicer.
  234. 9:03One important note: you only learn to
  235. 9:05build a dashboard by doing it, watching
  236. 9:07videos alone isn't enough. So, when
  237. 9:09should you move on to the next step?
  238. 9:11When you have created a one-page
  239. 9:13dashboard using Kaggle data that shows
  240. 9:15monthly sales and top-selling categories
  241. 9:17.
  242. 9:18Step five: Python. Now, why Python?
  243. 9:22There are two reasons.
  244. 9:24First, an Excel sheet can't hold more
  245. 9:26than about a million rows.
  246. 9:29When data gets larger, Excel falls short
  247. 9:31.
  248. 9:32The more important reason is that even
  249. 9:34when AI writes Python code for you, you
  250. 9:36need to be able to read and understand
  251. 9:38what it’s doing, though you don’t need
  252. 9:40to be a programmer. So, what exactly do
  253. 9:43you need to learn? In this order? First
  254. 9:45, Python basics: variables, lists, and
  255. 9:48conditions; all of that is enough.
  256. 9:51Then libraries—Python has a set of
  257. 9:53pre-made toolkits so you don’t have to
  258. 9:55sit there and write everything from
  259. 9:57scratch.
  260. 9:59These are called libraries.
  261. 10:01For data analysis, two of them are
  262. 10:03enough for you. Pandas is for working
  263. 10:05with tables, and NumPy is for working
  264. 10:07with numbers. Companies like Digikala
  265. 10:10and Digishahr mentioned these two
  266. 10:11libraries by name in their job postings
  267. 10:14. Finally, where should you write your
  268. 10:16code?
  269. 10:17There’s a notebook where you can write
  270. 10:18your code in pieces and see the result
  271. 10:20of each piece right there.
  272. 10:24These notebooks are called Jupyter
  273. 10:26Notebooks, and you can install them on
  274. 10:28your own system for free without needing
  275. 10:30a powerful computer.
  276. 10:35The online version is Google Colab,
  277. 10:37which we used, and it is currently
  278. 10:39accessible from Iran.
  279. 10:42Now, when should we move on to the next
  280. 10:44step?
  281. 10:45When you can recreate the pivot table
  282. 10:47from the first step—the number of orders
  283. 10:50each month—using just a few lines of
  284. 10:52Pandas. Step six: Artificial
  285. 10:54Intelligence. Now for the final step and
  286. 10:57the answer to the video's first question
  287. 10:59: don't just see AI as a line on your
  288. 11:01resume; it should become your daily
  289. 11:03assistant.
  290. 11:04So, what is it good for?
  291. 11:08Learning every stage of this data
  292. 11:10analysis path we are talking about.
  293. 11:14Writing the first version of a query.
  294. 11:16You give it your question in Persian,
  295. 11:18and it delivers a query; explaining
  296. 11:20errors—for example, if your Python code
  297. 11:23has an error, you give it the error and
  298. 11:25ask what it means. Where is the problem
  299. 11:27? Repetitive tasks!
  300. 11:30If you download a report every week,
  301. 11:32repeat the same cleaning, and keep
  302. 11:34sending it to your manager, ask which
  303. 11:36parts can be handed over to AI to build
  304. 11:38an automation.
  305. 11:42All these tasks we’ve been discussing
  306. 11:44are called automation.
  307. 11:46But one skill here is more important
  308. 11:48than all: knowing whether the answer the
  309. 11:51AI gave is actually correct or not.
  310. 11:54Let me give you a figure; there’s a
  311. 11:56famous test called "Bird" where they
  312. 11:58give AI questions to write SQL correctly
  313. 12:00.
  314. 12:02The best models, when working alone, got
  315. 12:05about 80 percent right, but human
  316. 12:07experts got about 93 percent right.
  317. 12:11Meaning, in that test, almost one out of
  318. 12:14every five queries the AI wrote was
  319. 12:16incorrect. Let me tell you a common
  320. 12:18mistake that happens with this very
  321. 12:20Kaggle data. You ask the AI, for example
  322. 12:22, to calculate total sales.
  323. 12:26It joins the payment table to the order
  324. 12:28items table.
  325. 12:32Now, if an order has 3 items, the amount
  326. 12:35for that order is counted three times,
  327. 12:37making sales higher than reality, and
  328. 12:39the scary part is the query runs without
  329. 12:42any errors.
  330. 12:45The only person who can catch this error
  331. 12:47is someone who knows SQL themselves.
  332. 12:50So, the answer to the first question of
  333. 12:52our video is exactly this. Do companies
  334. 12:54still want SQL? They want someone who
  335. 12:57knows SQL because someone needs to check
  336. 12:59the AI's answer to see if what it wrote
  337. 13:01is correct or not.
  338. 13:02Now, when do you learn this step?
  339. 13:05When you give the second-step question
  340. 13:07to the AI, compare its query with your
  341. 13:09own, and if they differ, understand
  342. 13:11which one is correct and why.
  343. 13:14The most important skill of all. Well,
  344. 13:16our 6 steps are finished.
  345. 13:19But there is a skill you must practice
  346. 13:21alongside all these steps, and AI will
  347. 13:23help you the least with this one.
  348. 13:27That is reaching a decision from numbers
  349. 13:29.
  350. 13:31SnappFood wrote in its delivery team ad
  351. 13:33that your job is to be able to find the
  352. 13:35root cause of major problems.
  353. 13:38Sedat also wrote that you should be able
  354. 13:40to explain complex ideas simply. These
  355. 13:42points have two parts.
  356. 13:46The first part, understanding the
  357. 13:48business itself; AI can say a number has
  358. 13:50dropped, but it doesn't know what
  359. 13:52happened outside the table, within the
  360. 13:54company, to cause that drop.
  361. 13:58To start, choose one sector and learn
  362. 14:01its key metrics.
  363. 14:05For example, if you like online retail,
  364. 14:07see how many out of every 100 people who
  365. 14:09open the site actually make a purchase;
  366. 14:12this number is called the conversion
  367. 14:14rate.
  368. 14:17The second part is storytelling.
  369. 14:20You shouldn't take the result of your
  370. 14:22analysis to your manager with 20 charts
  371. 14:24.
  372. 14:24Take 3 sentences.
  373. 14:26Let me give you an example using that
  374. 14:28same Kaggle data.
  375. 14:30Suppose in your analysis you saw that
  376. 14:32orders that arrived late received lower
  377. 14:34scores. You go to your manager and say,
  378. 14:36here is what I found. Orders that arrive
  379. 14:39late have lower customer ratings. Why is
  380. 14:41it important? Dissatisfied customers
  381. 14:43won't buy from us again. What is my
  382. 14:45recommendation?
  383. 14:48I suggest we start by solving the issues
  384. 14:50in cities with the most delays; use a
  385. 14:52chart, not 20—we found why it's
  386. 14:54important and what my proposal is.
  387. 14:59Now, the projects; you don't need 10,
  388. 15:01just 3 strong ones, and the good thing
  389. 15:04is if you've followed the steps with
  390. 15:06that Kaggle data I mentioned, you're
  391. 15:08already halfway there.
  392. 15:10Project one: A sales dashboard with
  393. 15:12Power BI. Project two: An analysis with
  394. 15:15SQL, for example, the best-selling
  395. 15:18categories and seasons. Project three:
  396. 15:21An analysis with Python that includes a
  397. 15:23suggestion at the end, like the
  398. 15:25relationship between delays and customer
  399. 15:27scores for the next step, or choose
  400. 15:29based on job ads.
  401. 15:31For instance, Snapp Market mentions in
  402. 15:34their ad that experience in e-commerce,
  403. 15:36fintech, or logistics is a plus.
  404. 15:39So, choose your projects in one of these
  405. 15:42areas.
  406. 15:43And let me tell you one more tip.
  407. 15:45Don't just dump your code on GitHub and
  408. 15:47think you're done.
  409. 15:49For each project, write down what the
  410. 15:51problem was, what you discovered, and
  411. 15:53what you suggest.
  412. 15:55Because employers want to see how you
  413. 15:57think, not just that you know how to
  414. 15:58code.
  415. 15:59Now, let's go to the last step. Finding
  416. 16:02a job.
  417. 16:03Don't wait until you've learned
  418. 16:04everything perfectly.
  419. 16:06Start sending out resumes as soon as you
  420. 16:08finish your second project. Don't
  421. 16:10underestimate internship ads either. For
  422. 16:12example, DigiShahr wrote for a data
  423. 16:14analyst intern that experience doesn't
  424. 16:16matter at all. Basalam also wrote that
  425. 16:18they are looking for everyone from
  426. 16:19interns to seniors. For the interview,
  427. 16:22practice these items.
  428. 16:25Have a practice and review of SQL and
  429. 16:27Excel questions, as they are usually the
  430. 16:29first filters.
  431. 16:32A business problem that you must solve
  432. 16:34right there in the interview; this is
  433. 16:36called a case study, which you should
  434. 16:38practice with AI.
  435. 16:41And behavioral questions about teamwork
  436. 16:43are usually asked.
  437. 16:48And do one more thing: ask AI to play
  438. 16:50the interviewer for you; tell it: "You
  439. 16:52are a data analyst interviewer, give me
  440. 16:55a case study, and critique and correct
  441. 16:57my answers."
  442. 17:00Let's summarize what you need to do.
  443. 17:03Okay, I'll summarize the whole path for
  444. 17:05you once. Excel is for getting familiar
  445. 17:08with data.
  446. 17:09SQL is a data analyst's most important
  447. 17:12skill.
  448. 17:13Basic statistics so you understand if a
  449. 17:15number is significant or not. Power BI
  450. 17:18for dashboards.
  451. 17:20Python for big data and AI as an
  452. 17:22assistant (provided you check its
  453. 17:24answers yourself), and alongside all of
  454. 17:26them is getting from numbers to
  455. 17:28decisions.
  456. 17:30If I had to tell you just one thing to
  457. 17:33start this week: download that Kaggle
  458. 17:35data, open it in Excel, and build your
  459. 17:37first pivot table.
  460. 17:40AI is changing the data analyst's job,
  461. 17:42but it won't eliminate it.
  462. 17:45Anyone who knows data, understands
  463. 17:47business, and uses artificial
  464. 17:49intelligence as an assistant, has a
  465. 17:51secure place in this market. I am very
  466. 17:53happy that you have joined me this far
  467. 17:55in the tutorial. If you would like to
  468. 17:57take the next step and see what the path
  469. 17:59to becoming an AI engineer is like, be
  470. 18:01sure to watch this video. Good luck, and
  471. 18:03I will see you in the next video.

About this transcript

This page contains the full transcript of یادگیری تحلیل داده از صفر با AI | بدون این ۶ قدم، Data Analyst نمیشی! by Hozhi, generated from the public captions YouTube serves with the video. The transcript has 2,853 words across 471 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.