mc_forecast 0.1.0 → 0.2.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 4f0d344256cf08700cd6d8475d0ac19ec44079cae9699f7a037fc1b934ee5c4d
-  data.tar.gz: dca662a52e4bc902ddf944e47e59882f43d86629cf5792aa19d737fa430f8d59
+  metadata.gz: 82e81b59b88ffa6eadba70aef437ecec5c040b890b19d7f64f8237eb59fc7c56
+  data.tar.gz: 729b9f9d931b338f5860427d1aaba73aea35c847eedb018d72c39a7486c7f551
 SHA512:
-  metadata.gz: a066871adf87c23610e0f08938934f5e1abdb630e477aea65e86195446952e279df54028c9ea12017e37669371203d6f22c53d21c1433aa69cf88c51ccb7b6b1
-  data.tar.gz: 3c294be97dc9040d7cde19ae90c0508d02fe2f6eb777cb38ff109ac7a7971ca94b25b1fbf47cd53190cdcb5eb2f2a3558198361ffa54d6d9fdd43b7004d63982
+  metadata.gz: 4ce13d962a1c9e82f52e101387f6c8627fdac52187000e5ceb5a25bc7513bc9929cd94d6fc6faaf5196f7bcc90336bb69be642c774abe93a205b43464bbc20f1
+  data.tar.gz: ea2d982df7f53a36558969c1d8361ea477d817895f92a4118cb23472eaa3066b85ce9efffcc12f804894facf1303f8dffe68c2b2bb9fd8865125d0ba516a41fd

data/README.md

@@ -5,7 +5,7 @@
 [![GitHub Workflow Status](https://img.shields.io/github/actions/workflow/status/DaanVanVugt/ruby-mc/ci.yml)](https://github.com/DaanVanVugt/ruby-mc/actions/workflows/ci.yml)
 [![Code Climate maintainability](https://img.shields.io/codeclimate/maintainability/DaanVanVugt/ruby-mc)](https://codeclimate.com/github/DaanVanVugt/ruby-mc)
 
-TODO: Description of this gem goes here.
+Use Monte-Carlo methods for business forecasting. Define transition methods (for example a month-based one) and keep track of events you are interested in. Automatically generates a 95% confidence interval and mean values.
 
 ---
 
@@ -23,6 +23,17 @@ gem install mc_forecast
 
 ```ruby
 require "mc_forecast"
+# all arguments optional
+e = McForecast::Simulation.new.run(init_state: nil, steps: 1, trials: 1_000) do |_state, _step, _trial|
+  events = {}
+  events[:coin] = rand > 0.5 ? 1 : 0
+
+  # block should return a new state and a hash of events
+  [nil, events]
+end
+# e[:coin][:mean][0] ~ 0.5
+# e[:coin][:quantiles][0.025][0] ~ 0
+# e[:coin][:quantiles][0.975][0] ~ 1
 ```
 
 ## Support

data/lib/mc_forecast/simulation.rb

@@ -1,9 +1,11 @@
+require "deep_dup"
+
 module McForecast
   class Simulation
-    def run(init_state: nil, trials: 1_000, steps: 1, quantiles: [0.025, 0.975])
+    def run(init_state: nil, trials: 1_000, steps: 1, quantiles: [0.025, 0.16, 0.84, 0.975])
       events = {}
       (0..trials - 1).each do |trial|
-        state = init_state.dup # this is not a deepcopy!
+        state = DeepDup.deep_dup(init_state)
         (0..steps - 1).each do |step|
           state, e = yield state, step, trial
           e.each_pair do |k, v|
@@ -22,7 +24,11 @@ module McForecast
 
     # Return an analysis of the events, containing:
     # { event_name:
-    #   { mean: [...], # per step
+    #   {
+    #     sum: { mean: ...,
+    #            quantiles:
+    #              { 0.025: ..., 0.975: ... }},
+    #     mean: [...], # per step
     #     quantiles:
     #      { 0.025: [...],
     #        0.975: [...]
@@ -30,20 +36,39 @@ module McForecast
     def analyze(events, quantiles)
       events.transform_values do |steps| # array(steps) of arrays(trials)
         {
-          mean: steps.map { |trials| Rational(trials.sum || 0, trials.length) },
-          quantiles: Array.new(steps.length)
+          sum: sum(steps, quantiles),
+          # besides the total sum we may want to have a sum for multiples of our base period
+          # (or week/month/quarter/year but that gets a bit complicated)
+          mean: steps.map { |trials| (trials.sum || 0).to_f / trials.length },
+          quantiles: quantiles.zip(step_quantiles(quantiles, steps)).to_h
         }
+      end
+    end
+
+    def sum(steps, quantiles)
+      sums = steps.transpose.map(&:sum) # gives a sum of this event, per trial
+      {
+        mean: (sums.sum || 0).to_f / steps[0].length,
+        # sort all of the sums, and take the elements closest to the chosen quantiles, and then make a nice hash
+        quantiles: quantiles.zip(sums.sort.values_at(*quantile_indices(steps[0].length, quantiles))).to_h
+      }
+    end
 
-        require 'byebug'
-        puts(steps.map do |trials|
-          [quantiles, trials.sort.values_at(*quantile_indices(trials.length, quantiles))]
-        end)
+    def step_quantiles(quantiles, steps)
+      if quantiles.any?
+        # only need to sort if we request answers on any quantiles
+        steps.map do |trials|
+          # could avoid sorting with some creativity, but probably fine for our data lengths so far
+          trials.sort.values_at(*quantile_indices(trials.length, quantiles))
+        end.transpose # a[step][]
+      else
+        []
       end
     end
 
-    def quantile_indices(n_trials, quantiles)
+    def quantile_indices(count, quantiles)
       quantiles.map do |q|
-        (q * (n_trials - 1)).round.to_i.clamp(0, n_trials - 1)
+        (q * (count - 1)).round.to_i.clamp(0, count - 1)
       end
     end
   end

data/lib/mc_forecast/version.rb

@@ -1,3 +1,3 @@
 module McForecast
-  VERSION = "0.1.0".freeze
+  VERSION = "0.2.0".freeze
 end

metadata

@@ -1,15 +1,29 @@
 --- !ruby/object:Gem::Specification
 name: mc_forecast
 version: !ruby/object:Gem::Version
-  version: 0.1.0
+  version: 0.2.0
 platform: ruby
 authors:
 - Daan van Vugt
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2023-12-08 00:00:00.000000000 Z
-dependencies: []
+date: 2024-03-19 00:00:00.000000000 Z
+dependencies:
+- !ruby/object:Gem::Dependency
+  name: deep_dup
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: '0'
+  type: :runtime
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: '0'
 description: 
 email:
 - dvanvugt@ignitioncomputing.com